KR20210119384A - Liquid Immersion Cooling Platform - Google Patents

Abstract

A two-phase liquid immersion cooling system is described in which a heat generating computer component causes a dielectric fluid to evaporate from its liquid phase. The dielectric vapor is then condensed back into the liquid phase and used to cool computer components. Using a pressure control vessel and a pressure controller, the disclosed system can be operated below ambient pressure. By controlling the pressure at which the system operates, the user can influence the temperature at which the dielectric fluid evaporates, thereby achieving increased performance from a given computer component. With the use of robotic arms and slot-in computing components, self-healing computing systems can be created.

Description

Liquid Immersion Cooling Platform

FIELD OF THE INVENTION The present invention relates to immersion cooled computing systems, ie, immersion cooled computing systems that utilize pressure and/or vapor management.

Typical computing and/or server systems use air to cool various components. Conventional liquid or water cooled computers use a flowing liquid to remove heat from the computer components, but avoid direct contact between the computer components and the liquid itself. The development of electrically non-conductive fluids and/or dielectric fluids is the use of immersion cooling to allow computer components and other electronics to be immersed in a dielectric or electrically non-conductive liquid to remove heat directly from the component into the liquid. makes it possible Immersion cooling can be used to reduce the total energy required to cool computer components, and can also reduce the amount of space and equipment required for proper cooling.

In the disclosed embodiments of the invention described below, the use of a steam and pressure management system as well as a power management system may be utilized, individually or in combination, to create a significantly improved computer system that utilizes liquid immersion cooling. have.

SUMMARY Embodiments of the disclosed invention are directed to a pressure controlled vessel that may be used to house a liquid immersion cooling computing system. In some embodiments, the pressure control vessel contains a liquid dielectric fluid in an amount sufficient to substantially submerge the heat generating computer component, and also contains an atmosphere comprising the gaseous dielectric fluid. Embodiments further include a condensing system for cooling and converting the gaseous dielectric fluid to a liquid dielectric fluid. The disclosed pressure management system allows the disclosed embodiments to be operated under vacuum, thereby evaporating the dielectric fluid and reducing the temperature at which the computing system operates. The disclosed embodiments enable an increase in the density and/or computing power of computer components due to the improved temperature management system described.

1 and 2 show schematic views of a pressure control vessel according to an exemplary embodiment.
3 shows the exterior of an exemplary embodiment of a pressure control vessel 110 .
4 shows an exemplary embodiment of a superstructure comprising a plurality of pressure control vessels.
5 depicts an exemplary data center embodiment showing multiple pressure control vessels connected to a central power source.
6 depicts an exemplary data center embodiment showing multiple pressure control vessels connected in series with each other.
7A-7D illustrate an exemplary embodiment of a cooled computing system having an inner robotic arm, an airlock, and an outer robotic arm.
8A-8C show an exemplary embodiment of a rack system.
9A-9G show exemplary embodiments of a chassis for mounting various components.
10A-10F show an exemplary embodiment of a pressure control vessel.
11 shows an exemplary cooling and vapor management system for a pressure control vessel.
12A-12E show another embodiment of a container.
13 shows an example of a stand-alone container.
14 shows an example of an outer housing for a free-standing container.
15A-15D show an exemplary magazine positioned on a platform that may extend out of the container.
16 shows a vapor recovery system according to an exemplary embodiment.
17 shows an exemplary embodiment of a rack power distribution system.
18 shows an example of a heating element for an immersion cooling system according to an exemplary embodiment.
19A and 19B show a filter comprising three cores according to an exemplary embodiment.
20A and 20B show an exemplary robotic system.
21A and 21B show an exemplary guide pin mechanism between the chassis and the rack.
22 depicts an exemplary connector with self-aligning features.

In the following description, specific details are set forth, such as specific quantities, sizes, arrangements, configurations, components, and the like, in order to provide a thorough understanding of the embodiments disclosed herein. However, it will be apparent to one skilled in the art that the present disclosure may be practiced without such specific details. In many cases, such considerations and other details may be omitted where such details are not necessary for a thorough understanding of the present disclosure and are included within the function of one of ordinary skill in the relevant art.

Equipment, components, systems, and subsystems of some disclosed embodiments are described below with reference to trade names. It is understood that the present disclosure may be practiced with many similar components, whether similar components are developed and/or sold under a particular trade name, and that the features and/or limitations associated with components of a particular trade name are not essential to the practice of the disclosed invention. It will be clear to those skilled in the art.

dielectric fluid

에 의한 엔지니어링된 유체의

계열의 일부를 포함하나, 설명된 발명은 어떠한 특정 유전성 유체로도 제한되지 않는다. 일부 침잠 유체는 전형적으로, 냉각되는 컴퓨터 구성 요소를 동작시키고자 하는 온도에서 비등점을 갖는다. 개시된 시스템의 다른 양태뿐만 아니라 모든 컴퓨터 구성 요소는 바람직하게, 유전성 유체와 접촉될 때 압력 제어 용기 내에서 용해되지 않고 달리 파괴되지 않는 재료로 제조된다. 일부 실시예에서, 표준 대기압에서의 유전성 유체의 비등점은 약 100 ℃ 미만, 또는 약 80 ℃ 미만, 또는 약 60 ℃ 미만, 또는 약 50 ℃ 미만, 또는 그 미만일 수 있다. 일부 실시예에서, 표준 대기압에서의 유전성 유체의 비등점은 약 60 ℃ 초과, 또는 약 40 ℃ 초과, 또는 약 30 ℃ 초과, 또는 약 20 ℃ 초과일 수 있다. 침잠 냉각 유체의 특정 실시예는 일반적으로 낮은 증기압을 갖는다. 침잠 냉각 유체의 일부 실시예는 플루오로카본 및/또는 불소화 케톤이다. 유전성 유체의 특정 실시예는 (CF3)2CFCF2OCH3, C4F9OCH3, 또는 CF3CF2CF2CF2OCH3의, 또는 그와 유사한 화학식을 가질 수 있다. 특정 유전성 유체는 하이드로플루오로 에테르, 메톡시-노나플루로부탄을 포함한다.One aspect of immersion cooling is the use of a thermally conductive, but electrically substantially non-conductive, or substantially dielectric fluid. Examples of such fluids include Novec 7100,

of engineered fluids by

While including part of the series, the described invention is not limited to any particular dielectric fluid. Some immersion fluids typically have a boiling point at the temperature at which they are intended to operate the computer component being cooled. All computer components, as well as other aspects of the disclosed system, are preferably made of materials that do not dissolve or otherwise break within the pressure control vessel when contacted with the dielectric fluid. In some embodiments, the boiling point of the dielectric fluid at standard atmospheric pressure may be less than about 100 °C, or less than about 80 °C, or less than about 60 °C, or less than about 50 °C, or less. In some embodiments, the boiling point of the dielectric fluid at standard atmospheric pressure may be greater than about 60°C, or greater than about 40°C, or greater than about 30°C, or greater than about 20°C. Certain embodiments of immersion cooling fluids generally have low vapor pressures. Some examples of immersion cooling fluids are fluorocarbons and/or fluorinated ketones. Certain embodiments of the dielectric fluid may have a formula of (CF3)2CFCF2OCH3, C4F9OCH3, or CF3CF2CF2CF2OCH3, or similar. Certain dielectric fluids include hydrofluoro ethers, methoxy-nonaflurobutane.

Other desirable properties of an immersion cooling fluid include low toxicity, nonflammability, and/or low surface tension. In some embodiments, the immersion cooling fluid does not substantially damage computer components and/or connections, wires, cables, seals, and/or adhesives associated with the computer components at the pressures and temperatures utilized for liquid immersion cooling. does not Some dielectric fluids have a dielectric constant in the range of about 1.8 to about 8 and a dielectric strength of about 15 megavolts per meter (15 MV/m). In some embodiments, the dielectric fluid has a dielectric strength of at least about 5 MV/m, or at least about 8 MV/m, or at least about 10 MV/m, or at least about 12 MV/m. In some embodiments, the dielectric fluid has a dielectric strength of about 3 MV/m or less, or about 5 MV/m or less, or about 8 MV/m or less. In the disclosed embodiment, any liquid in contact with the computer component 170 has a sufficiently high dielectric strength to prevent damage to the computer component at intervals and conditions of certain applications.

Some dielectric fluids have a critical heat flux of at least about 10 W/cm 2 , or at least about 15 W/cm 2 , or at least about 18 W/cm 2 , or at least about 20 W/cm 2 . Some dielectric fluids have a critical heat flux of about 15 W/cm2 or less, or about 10 W/cm2 or less, or about 8 W/cm2 or less, or about 5 W/cm2 or less.

1 shows a schematic diagram of a cooled computing system 110 in accordance with an exemplary embodiment. Embodiments of the disclosed cooled computing system 110 (or computing system, system, vessel, or pressure control vessel, all of which may be used interchangeably) provide for submerging a component into a bath of fluid. may use the liquid dielectric fluid 140 to cool the computer component 170 . As electricity passes through component 170 , component 170 generates heat. As component 170 heats up, the performance of the component may decrease or the component may be damaged to a point of failure. It is advantageous to keep several computing components stable and at relatively low temperatures. In some embodiments, computer component 170 may be maintained at less than about 80 °C, or less than about 70 °C, or less than about 65 °C, or less than about 60 °C, or less than about 55 °C. In some embodiments, computer component 170 can be maintained above about 60°C, or above about 50°C, or above about 40°C, or above about 35°C, or above about 30°C. As the computer component 170 is heated, heat is transferred to the liquid dielectric fluid 140 surrounding the component 170 . When the liquid dielectric fluid reaches its boiling point, it will convert from the liquid phase to the gas phase and rise out of the liquid bath 142 . The components 170 within the bath 142 of dielectric fluid can generally be maintained around the boiling point of the particular dielectric fluid 140 used.

When the liquid dielectric fluid, at the pressure used for a given application, is heated to its vaporization point and begins to become a gas, bubbles of dielectric vapor will rise out of the liquid bath 142 and rise to the top of the system 110 . something to do. The vapor is then cooled to the condensing point using a condenser 130 . Depending on the configuration of the system 110 , heating and cooling of the dielectric fluid from the liquid phase to the vapor phase and vice versa can create a convection current as shown in FIG. 2 .

In some embodiments, when the system is operated, the computer component 170 will be completely submerged in the liquid dielectric fluid 140 . In other words, the upper portion of the computer component 170 is located below the level of the dielectric liquid 140 . It will be appreciated that small bubbles of dielectric fluid vapor contact the computer component as heat from the computer component causes the dielectric fluid to change from a liquid phase to a gas phase. Such a component would still be considered entirely submerged in the liquid phase of the dielectric fluid. In some embodiments, computer component 170 may be submerged in a liquid phase of dielectric fluid 140 . In one exemplary embodiment, any portion of computer components including, but not limited to, a motherboard, chip, server, card, blade, GPU or any portion of a CPU, and/or any peripheral component is hereditary. A computer component will be considered submerged when in direct contact with the liquid phase of fluid 140 . In certain embodiments, computer component 170 may be at least partially immersed in a liquid phase of dielectric fluid 140 . If the computer component 170 is not submerged, but is sufficiently cooled by the dielectric vapor, the computer component will be considered at least partially submerged.

In some existing immersion cooling systems, the dielectric fluid must be continuously added to the bath of dielectric fluid because the fluid is constantly boiling. Failure to add dielectric fluid to bath 142 may result in a low level of dielectric fluid in bath 142 until the components are exposed to a gaseous atmosphere and not cooled properly. This may result in reduced performance or damage to component 170 .

In some embodiments, there may be multiple modes of operation that may be described as fluid management systems involving dielectric fluids in a liquid state. These modes include (1) initial filling, which is a process by which a dielectric fluid is transferred from a storage system into a container; (2) continuous leveling, a process that allows additive fluid to be added to the vessel or excess fluid to be removed from the vessel; (3) Unfilling, a process that allows fluid to drain from the container and to be placed into a storage system; and (4) operational filtering, a process that allows the fluid to be continuously circulated through the filtering system to ensure removal of any particulates.

In some embodiments, the first three liquid management goals: initial filling, continuous leveling and emptying may be achieved through the same full set of piping, pumps and valves. A dedicated tank for storing liquid coolant can be used for storage of fresh fluid and excess fluid that is removed and re-condensed during the vapor management process. A set of pipes and pumps can be used to move coolant (or dielectric fluid) from the storage system to the vessel during filling and leveling, and to the outside of the vessel and back into the storage system during emptying operations.

In some embodiments, the fourth of the liquid management goals, operational filtering may be achieved through a series of skimmers and/or filters. The first stage may be a large particle filter positioned within the lower end of the vessel. The purpose of this filter is to prevent particles that are too large to be handled at a later stage from entering the rest of the system. The second stage may be a median particulate filter disposed in-line within the piping system between the first stage and the third stage. This second stage intermediate particulate filter may utilize a small barrel style filter to remove particulates that are too small to be removed by the first stage filter but too large to be handled by the third stage filter. . The third stage filter may be composed of one or more parallel filters supporting various types of filter configurations. In some embodiments, a particular style of filter will be defined by performing an analysis of the fluid after the fluid has been exposed to and operated with a set of hardware components positioned within the container environment. Different hardware and/or components may create different types of particulates and chemicals that may need to be filtered to ensure long life and efficiency of the dielectric fluid.

pressure management

In general, the immersion cooling fluid should be kept free of dust, water, and/or other contaminants. As computer component 170 is in direct contact with immersion cooling fluid 140 , minor contaminants can cause shorting or damage to the computer component. Also, water or water vapor, which can contaminate the dielectric fluid, can reduce the dielectric properties of the fluid, including but not limited to dielectric strength, when the fluid begins to become contaminated. When the dielectric strength of the dielectric fluid is reduced, computer components may be shorted or otherwise damaged during operation. One way to reduce contamination is to operate an immersion cooling system in an enclosure that is maintained at or slightly above atmospheric pressure.

When the computer component 170 is operated, the heat generated from initial use of the computer component causes some dielectric liquid 140 to evaporate into a gas. Where the immersion cooling system is confined within a substantially enclosed housing, this evaporation typically increases the pressure of the atmosphere within the housing. Pressure relief valves, expansion sheaths, and/or other techniques may be used to limit pressure build-up, and/or to maintain the pressure within the housing at or only slightly above atmospheric pressure. Maintaining some positive pressure within the enclosure can help reduce the penetration of dust, water vapor, or other contaminants into the submerged cooling computing system.

The present embodiment provides an enclosed pressure control vessel 110 (or cooling) for housing computing components 170 and immersion cooling equipment, as well as associated power, networking connections, wiring connections, and others within the pressure control vessel. The computing system 110 that becomes the enclosure is used. In contrast to existing models, pressure controlled vessel 110 can be maintained at at least some vacuum, thereby reducing the boiling point of dielectric fluid 140 to a temperature below its boiling point at standard atmospheric pressure.

By operating the computing and immersion cooling system under vacuum, component 170 can be maintained at the reduced, low pressure boiling point of dielectric fluid 140 . This has the benefit of increased cooling, which may allow more electricity to pass through the various components 170 resulting in higher performance of the components. By controlling the pressure in the pressure control vessel 110 , the boiling point of the dielectric fluid 140 can also be controlled, thereby allowing the same fluid 140 to be used in a wider range of conditions. Many embodiments benefit from cooler temperatures, but certain computer components 170 may have ideal ranges, and performance may decrease at temperatures below that range. By controlling the pressure in the pressure control vessel 110 , the boiling point of the immersion cooling fluid 140 may also be controlled. In certain embodiments, the disclosed pressure management system may be used to dynamically control the pressure, and thereby the boiling point of the dielectric fluid 140 , when the computing system is started, stopped, or in response to other condition changes. have.

In addition to reducing the boiling point of dielectric fluid 140 by operation within pressure controlled vessel 110 below ambient pressure, more efficiently transfer heat away from computer component 170 itself and into dielectric fluid 140 . To this end, the computer component 170 itself may be modified. By increasing the surface area of a component 170 , eg, a chip, exposed to the liquid dielectric fluid 140 , heat transfer between the component 170 and the bath 142 of the dielectric fluid 140 is increased. can be An exemplary apparatus for increasing the surface area may be a copper boiler or copper disk that may be attached to a chip of another computer component 170 . In a particular embodiment, the bond being used will be selected based on its ability to transfer heat and its solubility in the dielectric cooling fluid. Preferred adhesives exhibit high thermal conductivity and low solubility in the selected dielectric fluid.

1 depicts a schematic diagram of an exemplary embodiment of the disclosed computing system. An embodiment of the disclosed system is an immersion cooling system comprising a pressure control vessel 110 (or computing system 110 to be cooled), a pressure controller 150 , at least a volume of dielectric fluid 140 and a condensing structure 130 . , and a desired computer component 170 . The pressure system may be configured to maintain a desired degree of pressure reduction. The pressure control vessel 110 maintains a negative pressure while still allowing multiple penetrations into the pressure control vessel 110 for various connections including, but not limited to, power, data, networking, cooling water, and/or communication systems. can be configured to Some embodiments use sealed and/or marine grade connections. Operating the computer system within the pressure-controlled vessel 110 below ambient pressure requires a series of modifications to the system as a whole. Such modifications are described below, and some are readily understood by those skilled in the art.

3 shows the exterior of an exemplary embodiment of a pressure control vessel 110 . In some embodiments, the disclosed pressure control vessel 110 is at least about 2 feet high, or at least about 3 feet high, or at least about 4 feet high, or at least about 5 feet high. In some embodiments, the pressure control vessel is no more than about 3 feet high, or no more than about 4 feet high, or no more than about 5 feet high.

In certain embodiments, the pressure control vessel is at least about 100 cubic feet, or at least about 150 cubic feet, or at least about 200 cubic feet, or at least about 250 cubic feet, or at least about 300 cubic feet, or at least about 350 cubic feet; or at least about 400 cubic feet of inner volume.

In some embodiments, the pressure control vessel will be configured to contain, in operation, about 12 vertical inches of liquid dielectric fluid and about 36 vertical inches of dielectric fluid vapor. In certain embodiments, the ratio of liquid volume to gas volume helps to create convection and direct the dielectric vapor towards the condensing structure, which converts the vapor back to a liquid. In some embodiments, the pressure control vessel is configured to include, during operation, a ratio of the volume of the liquid dielectric fluid to the volume of the gaseous dielectric fluid of about 1:6. In another embodiment, the pressure control vessel, during operation, is about 1:3, or about 1:5, or about 1:8, or about 1:10, or about 1:15, volume of liquid dielectric fluid to gaseous dielectric fluid. is constructed to contain the proportion of the volume of

In one exemplary embodiment, the pressure management system may include a pressure controller 150 . The pressure controller 150 may be a vacuum source, for example the pressure controller 150 may be a vacuum pump that may be connected to the pressure control vessel 110 . In some embodiments, the vacuum pump 150 may be located remotely and the vacuum may be delivered to the pressure control vessel 110 using piping and/or tubing. In a preferred embodiment, a pressure sensor 180 is housed within the pressure control vessel 110 and is used to regulate and/or maintain a desired negative pressure within the pressure control vessel 110 . In some embodiments, a pressure sensor 180 and/or a pressure regulator 190 is coupled to the processor, which uses the pressure sensor 180 to monitor the pressure in the pressure control vessel 110 and the pressure regulator 190 . Use to control the pressure.

Some embodiments include an operator protection mechanism. In one exemplary embodiment, the operator protection mechanism may be a locking mechanism that prevents the system from operating when any cover or service panel to the pressure control vessel is not in place. In one exemplary embodiment, the operator protection mechanism may include a controller for immediately shutting down power to the system in case of unauthorized tampering with one of the doors or panels of the pressure control vessel. In addition to providing life safety features, the operator protection mechanism may also provide enhanced operational security features for deployments where sensitive data is contained within the container. A high level of assurance in the effectiveness of the disk protection mechanism can be achieved by ensuring that the equipment is inaccessible during normal operation without powering down the system. Also, in some embodiments, the disk protection mechanism may use a runtime stored cryptographic key to protect data located in the pressure control vessel.

In certain embodiments, in addition to denying unsafe access to the pressure control vessel, sensors may be deployed to ensure that the system operates as designed. The primary sensor package includes a temperature sensor in the vapor space; a temperature sensor in the liquid space; humidity sensor in the vapor space; and/or a pressure sensor in the vapor space. Readings of these sensors may be monitored by software and/or human operators to ensure that the system operates in a safe and accurate manner. In some embodiments, sensor data will be recorded or analyzed later.

In some embodiments, additional sensors may be integrated into the container or a parent structure (defined below). Such sensors include, for example, FLIR-based thermal imaging cameras; VESDA or other types of smoke aspiration detectors; and/or a refrigerant leak detector designed to detect a leak of the dielectric fluid into the surrounding environment.

In some embodiments, the vessel and/or the upper structure may be provided with indicator lights related to the operating status of the system.

Although the cooled computing system 110 is often referred to as the pressure controller system 110 , one of ordinary skill in the art will appreciate that many or all of the advantages of the cooled computing system 110 may be realized without utilizing a “pressure control system”. will understand

steam management system

Liquid immersion cooling systems can be operated in different ways. Some may be operated by directly and continuously cooling the immersion fluid. Others may be operated by allowing the liquid to reach its maximum liquid phase temperature and then boiling into the vapor phase. An immersion cooling system operated by allowing a liquid to evaporate is referred to as a two-phase immersion cooling system. Two-phase immersion cooling systems often allow the dielectric fluid to boil and/or evaporate and add additive fluid regularly to replace lost fluid into the atmosphere.

The disclosed embodiment utilizes a liquid immersion cooling system housed within the pressure control vessel 110 . This has the advantage that there is no loss of the dielectric fluid 140 even after the dielectric fluid is converted to a gaseous form. Within the closed or substantially closed pressure control vessel 110 , a gaseous dielectric fluid may be condensed and a bath 142 of liquid dielectric fluid 140 that is actively used to cool the computing component 170 . can be added back to The condensing step may be effected in any convenient manner, for example by passing the process water through a thermally conductive tube. The condensing structure 130 may include radiator fins and/or similar equipment to increase the surface area of the condenser and thereby allow more and/or faster condensation of the gaseous dielectric fluid and return to liquid form. can In some embodiments, the process water is at ambient temperature and is not actively cooled. In other embodiments, the process water may be quenched using evaporative cooling, dry cooling towers, and/or other methods of chilling process water known in the art.

In some embodiments, there may be two interfaces between the pressure control vessel and the external system. The first interface may be a process water supply interface. This may be a pipe that carries the process water from the facility providing the quenched process water to a distribution manifold on the pressure control vessel. The second interface may be a process water return interface. This may be a pipe returning the process water to the facility providing the quenched water. The process water may be returned to the facility after the process water has flowed through the pressure control vessel and associated cooling component. The cooling component may include, for example, a condenser, a condensing coil, and/or a radiator within the vessel, as well as any powered component (including, for example, a motor, a pump, and/or a utility cabinet). It may include a coil that removes heat from the powered component). In some embodiments, there may be two interfaces between the upper structure and the external system. Such an interface may be similar or substantially similar to the two interfaces between the pressure control vessel and the external system.

In some embodiments, the location of the condensing structure 130 within the pressure control vessel 110 may be configured to optimize the flow of the vapor phase dielectric fluid and to increase the rate and/or efficiency of the condensation. In some embodiments, the geometry of the pressure control vessel 110 itself may be controlled to increase the rate and/or efficiency of condensation.

In one exemplary embodiment, the location of the condensing structure 130 allows for placement (eg, by a robot) of the computer component 170 within the vessel (or removal of the computer component 170 from the vessel). can be facilitated and optimized. For example, the condensing structure 130 may be disposed on the side (or sidewall) of the vessel such that the condensing structure 130 is not located between the lid of the vessel and the computer component 170 . Thus, when the lid is opened, the robot can directly remove the computer 170 without any interference with the condensing structure 130 . This arrangement of condensing structures may streamline the placement and removal of computer components 170, thereby providing significant advantages in autonomous operation of the vessel. In one exemplary embodiment, the condensing structure 130 may be located on a shelf within the vessel.

유전성 유체(140)를 이용하여 압력 제어 용기(110) 내에서 생성될 수 있다. 이는, 압력 제어 용기의 하단부에서 침잠 냉각 탱크 내의 약 12 인치 깊이의 액체 유전성 유체의 층을 남기는 한편, 압력 제어 용기 부피의 대부분은 기체이다. 압력 제어 용기의 천장은 길이방향으로 연장되는 구조물의 중간에서 더 낮다. 천장 및/또는 덮개(120)는 위쪽으로 각도를 이루고, 압력 제어 용기(110)의 측벽에 접근할수록 상승된다. 응축 구조물(130)은 압력 제어 용기(110)의 2개의 측면 상에서 길이방향으로 연장된다. 이러한 예시적인 실시예에서 응축 구조물(130)은 폭이 약 12 인치이고 높이가 약 24 인치이며, 실질적으로 압력 제어 용기(110)의 전체 길이로 연장된다. 응축 구조물(130)은, 유동하는 프로세스 물을 이용하여 냉각되는 큰 표면적의 핀들을 갖춘 라디에이터 유사 재료를 포함한다. 일부 실시예는 부가적으로 또는 대안적으로 열 교환기를 포함할 수 있다.1-3 , in one exemplary embodiment, the pressure control vessel is about 10 feet long, about 4 feet wide, and about 4 feet high. The bath 142 is approximately 130 gallons.

It can be created in the pressure control vessel 110 using the dielectric fluid 140 . This leaves a layer of liquid dielectric fluid about 12 inches deep in the immersion cooling tank at the bottom of the pressure control vessel, while the majority of the pressure control vessel volume is gaseous. The ceiling of the pressure control vessel is lower in the middle of the longitudinally extending structure. The ceiling and/or cover 120 is angled upward and rises as it approaches the sidewall of the pressure control vessel 110 . The condensing structure 130 extends longitudinally on two sides of the pressure control vessel 110 . In this exemplary embodiment, the condensing structure 130 is about 12 inches wide and about 24 inches high, and extends substantially the entire length of the pressure control vessel 110 . The condensing structure 130 comprises a radiator-like material with large surface area fins that are cooled using flowing process water. Some embodiments may additionally or alternatively include a heat exchanger.

As shown in FIG. 2 , the structural arrangement within the pressure control vessel 110 directs a convective flow of the dielectric fluid vapor as it rises from the liquid bath 142 after boiling. The structural arrangement directs the convective flow upwards towards the ceiling of the pressure control vessel, where the flow is directed towards the large surface area of the condensing structure 130 and condensed back into liquid form. The dielectric fluid 140 is then flowed back into the liquid bath 142 . In this way, the total amount of dielectric fluid 140 may be conserved within this closed housing. Using convection to circulate the dielectric fluid vapor allows the disclosed embodiments to be operated without a mechanical pump to circulate the dielectric fluid, thereby reducing the total energy use of the disclosed system.

Certain embodiments are advantageous when the pressure control vessel must be opened and/or during startup and/or shutdown of the system to enable redundant and robust control of the level of the liquid dielectric fluid. Additional tanks and/or storage containers of dielectric fluid that may be utilized may be utilized.

11 shows an exemplary cooling and vapor management system 600 for a pressure control vessel 110 . In this exemplary embodiment, the cooling and vapor management system 600 may include a quenched process water reservoir 611 extending through the cooling coil 132 to cause condensation of the dielectric fluid 140 . have. After passing through the cooling coil 132 , the process water may proceed to a process water return reservoir 612 . Cooling and vapor management system 600 may also include a tank for vapor storage 614 and a tank for dielectric fluid storage 615 . Tanks 614 and 615 may provide dielectric fluid or vapor when needed, for example during startup and/or shutdown of the system. In one exemplary embodiment, tanks 614 and 615 may be coupled via a condensing structure 616 . In the event of an oversupply of vapor in tank 614 , condensing structure 616 can remove the vapor and add it to fluid storage tank 615 as a dielectric fluid.

In some embodiments, during operation, the pressure control vessel is maintained at about 3 psi below ambient atmospheric pressure, which helps to lower the boiling point of the dielectric fluid and thereby lower the operating temperature of computer chips and other components. In some embodiments, the pressure control vessel 110 is at least about 2 psi below ambient pressure, or at least about 4 psi below ambient pressure, or at least about 6 psi, or at least about 8 psi, or at least about 10 psi below ambient pressure. maintain.

In some embodiments, it will be necessary to select components with some tolerance for pressure fluctuations. By adjusting the operating pressure of the system, it may be desirable to use a component capable of withstanding a wide degree of pressure in order to be able to manipulate the coolant boiling point and thus the general operating temperature of the overall system. Given the operating characteristics of a two-phase system, standard operating conditions for some embodiments may exhibit variations between ±4 PSIg. Under certain conditions, such as during rapid start-up or shutdown of the system, an additional PSIg difference of 3 may occur. In some embodiments, system level adjustments may be made to better control these variables and keep them within a more controlled and defined range.

In certain embodiments, computer component 170 operates at a pressure that is at least about 3% below ambient pressure, or at least about 5% below ambient pressure, or at least about 10%, or at least about 15%, or at least about 15% below ambient pressure. It is operated at a pressure as low as about 20%, or by at least about 25%, or by at least about 30%.

In some embodiments, the pressure control vessel, in operation, is less than about 750 torr, or less than about 710 torr, or less than about 650 torr, or less than about 600 torr, or less than about 550 torr, or less than about 500 torr, or about It is maintained at less than 450 torr, or less than about 400 torr. In some embodiments, the pressure control vessel, in operation, is greater than about 650 torr, or greater than about 600 torr, or greater than about 550 torr, or greater than about 500 torr, or greater than about 450 torr, or greater than about 400 torr, or about maintained above 300 torr.

Some embodiments use a vapor scrubbing process and/or an initial purging process to control the gas atmosphere within the pressure control vessel. This process removes a portion of the gaseous atmosphere from the pressure controlled vessel and removes undesirable portions of the atmosphere such as air and water vapor. These and other undesirable parts of the atmosphere can be separated based on the temperature at which the vapor condenses into a liquid. Due to the special properties and boiling point of dielectric fluids, many naturally occurring contaminants can be removed using this method. Removal of fluids that cannot readily condense helps maintain the purity of the dielectric fluid. A fluid will not be considered readily condensable if the condensation point of the fluid at standard atmospheric pressure is less than about 20° C. higher than the condensation point of the dielectric fluid, or if the condensation point of the fluid at standard atmospheric pressure is less than 10° C.

During maintenance, start-up and/or shutdown operations, to reduce the amount of dielectric fluid lost when the pressure control vessel is opened and/or exposed to atmospheric conditions, a blanket of inert gas, such as nitrogen gas, is may be introduced into the control vessel. As shown in FIG. 11 , the cooling and vapor management system 600 may include an inert gas tank 613 , which may supply an inert gas to reduce dielectric fluid loss.

Some disclosed embodiments may include a server and/or computing system that is substantially standalone. In some embodiments, special seals and/or connections may be used to reduce the total number of penetrations into the pressure control vessel 110 . Some embodiments combine power, water, vacuum, and networking connections into a bundle of lines, thus minimizing penetration into the pressure control vessel, reducing the potential for leakage while the system is under vacuum.

4 shows an exemplary embodiment of a superstructure comprising a plurality of pressure control vessels. In this exemplary embodiment, two pressure control vessels 110 are pre-piped, pre-wired, and housed within the modular signal superstructure 210 . This allows the embodiment to be pre-fabricated and delivered as a substantially complete standalone system. The modular system may be configured to connect to other modular embodiments of the disclosed computing system. In some embodiments, the modular superstructure 210 will require only a single power connection and will be pre-wired with appropriate electronics to supply the required voltages to the computer components and/or other electronic components.

5 depicts an exemplary data center embodiment showing multiple pressure control vessels connected to a central power source. 6 depicts an exemplary data center embodiment showing multiple pressure control vessels connected in series with each other. In this exemplary embodiment, the pressure control vessel 110 may or may not be disposed within the upper structure.

7A-7D illustrate an exemplary embodiment of a cooled computing system having an inner robotic arm, an airlock, and an outer robotic arm. In this exemplary embodiment, the internal robotic arm 230 housed within the pressure control vessel 110 may be used to remove the component 170 and deliver the removed component to the airlock 220 . By using the airlock 220 , the component 170 may be removed without substantially disrupting or not disrupting the pressure, atmosphere, dielectric fluid, and/or other conditions within the pressure control vessel 110 . have. Once component 170 is removed from pressure control vessel 110 , a replacement component may be introduced into pressure control vessel 110 using airlock 220 . The replacement component may then be installed by the inner robotic arm 230 . The use of components that can be installed in a “slot-in” manner, such as blade servers and chassis, can greatly facilitate this process.

A disruption of a condition within the pressure control vessel may be detected by means of a sensor, for example a pressure sensor, disposed within the pressure control vessel. Such collapse may be indicated by a condition that deviates from the standard range of operating conditions by at least 10%. A significant disruption of conditions within the pressure control vessel may be indicated by a condition that deviates from the standard range of operating conditions by at least 30%.

In certain embodiments, a standalone diagnostic program may be run that analyzes the performance of components within the pressure control vessel 110 . If component 170 does not perform as desired, robotic arm 230 may be used to automatically remove and/or replace component. In this manner, a self-healing, standalone server and/or computing system may be created. In certain embodiments, such self-healing systems include modular units that can be transported or delivered to remote locations using conventional methods to provide highly efficient computing power that requires limited setup and/or maintenance. To create, it can be pre-fabricated and pre-wired.

In some embodiments, the first vapor management goal for cooling the vapor and for allowing the vapor to condense back from the gaseous state to the liquid state is achieved entirely within the closed system of the vessel through the use of a condensing coil. The process water will be piped through a condensing coil in the vessel. The shape and geometry of the vessel itself will facilitate the flow of vapor from the bath region to the coil region, and gravity will serve to draw the re-condensed liquid back into the bath region.

In some embodiments, a second vapor management goal for monitoring and maintaining the internal pressure of the vessel is achieved through the use of an integrated pressure sensor within the vessel and the use of a purge system. In some embodiments, a purge system will be used to remove excess vapor from the vessel and condense it back into a liquid for storage in a liquid storage tank.

In some embodiments, the third vapor management goal for controlling and eliminating non-condensable components of vapor present during system startup is achieved through the same mechanism as the second goal. A purge system may be used during initial startup of the system to underpressure the system and to remove any non-condensable gases from the system.

In some embodiments, the fourth vapor management goal for controlling overlay of inert gas may be achieved using a dedicated nitrogen overlay supply system. This overlay keeps the coolant below the top of the container, thus minimizing the loss of coolant during the period the container is opened to service the components therein. Dedicated piping from a set of nitrogen storage tanks through a set of dedicated overlay pipes in the vessel will allow for the addition of an inert overlay when an operator wishes to open the system. This gas, along with any other non-condensables, may be removed during a non-condensable removal process, which may be carried out at system startup. The overlay vapor management process may be managed and monitored through the control system software, based on user commands and system health monitoring.

Ballast Block

In some embodiments of the disclosed system, such as that shown in FIG. 1 , the pressure control vessel 110 includes a deeper bath portion 142 for receiving a majority of the dielectric fluid 140 and a wider shelf area adjacent the bath ( 112) may be included. Boards, cards, chips, blades, and/or any other computer components 170 are received substantially within the deeper bath section 142 of the pressure control vessel 110 . The wider shelf area 112 may also contain the liquid dielectric fluid 140 and/or collect the re-condensed dielectric fluid 140 from the vapor phase to the liquid phase. In certain embodiments, the depth of the dielectric liquid within the pressure control vessel 110 may be increased using the ballast block 160 . The ballast block 160 can be used to occupy an undesirable volume on the shelf, thereby displacing any dielectric fluid 140 that may be on the shelf 112 , and additional dielectric fluid 140 . ) can be added to increase the level of the liquid. In some embodiments, the ballast block 160 includes riser feet 161 , which allow a fluid to flow under the ballast block 160 , such that the flow Allows condensed liquid to continue to flow into the deeper bath portion of the pressure control vessel, unobstructed by block 160 .

The ballast block 160 may be made of any material that does not interfere with the operation of the disclosed submerged cooling system. The ballast blocks may be made of materials including, but not limited to, metals, rubbers, silicones and/or polymers. Preferred materials are substantially insoluble in the dielectric fluid. The block should be denser than the dielectric fluid, but need not be solid. In a preferred embodiment, the block will have a handle or cut out which allows the block to be more easily handled and manipulated. Some embodiments of ballast block 160 utilize interlocking top and bottom sections, allowing the blocks to be stacked on top of each other in a reliable manner. The interlocking top and bottom portions reduce the risk that the block will damage any nearby components if the block is slid or otherwise displaced from its intended position. In some embodiments, the interlocking top is configured such that blocks can be securely stacked on top of the lowest block so that the lowest block does not impede fluid flow and occupy a significant volume, thereby allowing a significant amount It includes a recessed portion aligned with the pit and/or riser on the bottom portion so that the level of the dielectric liquid can be raised without the need to add additional dielectric liquid.

In some embodiments, the ballast block 160 is configured to extend the entire length of the pressure control vessel 110 and/or shelf 112 . In other embodiments, the ballast block 160 may have substantially any size that allows the block to be handled. In such embodiments, multiple modular ballast blocks may be configured to displace large or small volumes as desired. In some embodiments, a single ballast block is about 2 feet long or about 3 feet long or about 4 feet long or more, and about 6 inches wide, or about 8 inches wide, or about 12 inches wide, or more, and It has an external dimension of about 1 inch high, or about 3 inches high, or about 6 inches high, or about 8 inches high, or more.

upper structure

The disclosed computing system is comprised of several components, all of which may be attached directly or indirectly to a physical upper structure 210 , as shown in FIG. 4 . The upper structure 210 enables pre-wired and pre-piped connection of any desired electrical system, sensor system, control system, power system, fluid control system, pressure control system, and/or communication system. . This allows for faster and simpler deployment in the field, and allows testing in the factory before delivery to the customer.

The upper structure 210 is typically made of metal components and may be a skid mounted or configured to be handled by a forklift, hoist, or crane. In some embodiments, the upper structure 210 is configured to fit within a standard container to aid transport. The upper structure 210 and associated components may be configured to weigh less than about 58,000 lbs as a whole, and may be divided into smaller sub-components to help transport without special equipment. In some embodiments, the upper structure 210 and associated components will have a weight of less than about 50,000 lbs, or less than about 40,000 lbs, or less than about 30,000 lbs, or less than about 20,000 lbs. In some embodiments, the upper structure 210 and associated components will have a weight greater than about 5,000 lbs, or greater than about 10,000 lbs, or greater than about 20,000 lbs, or greater than about 30,000 lbs. Embodiments of the upper structure 210 may have any size and/or shape. Many embodiments are large enough to accommodate multiple pressure control vessels 110 , server racks 310 , and associated liquid immersion cooling equipment, as well as the necessary equipment for managing power delivery and distribution and network connectivity.

The overall design of the upper structure 210 may be adjusted to accommodate the unique aspects of each deployment, including customizing the types and amounts of power and process water interconnects to fit the needs of existing facilities.

Control and management systems for all components within the disclosed pressure control vessel may be included as part of the disclosed computing system. Preferred embodiments of the disclosed system include all mechanical systems necessary to maintain and operate a two-phase liquid immersion cooling environment, including the necessary pumps, valves, regulators, vapor management systems, pressure management systems, and other associated components. do.

The upper structure 210 may be of an open frame design, or may include side panels and access doors. This allows for deployment within or outside existing structures within the field location. The upper structure 210 may be modified to include weatherization features and thus to be deployed in harsh environments. In some embodiments, the upper structure may be a skid/module framework.

Various systems, features, and/or capabilities may be integrated into the upper structure 210 to support, monitor, and manage other components of the pressure control vessel and any environment contained within or associated with the pressure control vessel. In some embodiments, such systems may include fire detection and/or containment capabilities, dedicated air conditioning and/or environment management capabilities, security features such as access control, and/or monitoring features, among many others.

power system

Some embodiments of the upper structure 210 are designed to accommodate various electrical input means and to connect them to existing power distribution systems installed within the upper structure. One of the many exemplary embodiments includes a 415V input to the main breaker, which in turn distributes the 415V AC input to a series of power shelves that convert it to a 12V DC output. In a preferred embodiment, this conversion is substantially done in one conversion step, thereby reducing the lossy efficiencies typically associated with such conversion. A typical computer server location typically converts incoming industrial electricity from a high AC voltage such as 415V to a reduced AC voltage such as 120V. This conversion results in energy loss as heat. Under normal circumstances, this can result in an energy loss of about 6%. The 120V electricity must then be further converted into DC current for use by various computer components. This second conversion results in a second, about 6% energy loss to heat. By directly converting about 415V of industrial electricity to about 12V DC, the overall heat energy loss can be reduced.

Another example implementation may include a connection of a 480V AC input to a power shelf, which converts the 480V AC input to a 48V DC output, which in turn converts the 48V DC input to, for example, 12V, 5V. It is distributed to a series of intermediate power supplies that convert to various DC outputs including , 3.5V, 3.3V and others.

In some implementations, there may be one set of power supplies, or there may be multiple power supplies operating at different input and output voltages. The exact configuration will be tailored to the needs of the particular equipment being installed and to the conditions of the application. The specific design of the power system may be adjusted to suit the needs of the specific environment in which the disclosed computing system is deployed. Customization may include types, capacities, and interfaces for both input and output of power to the system.

In some embodiments, a rack power distribution system may include a modular power system and/or a set of modular power systems. The specific configuration of the modular power system or systems is not particularly important, as long as it can deliver the desired amount and type of power to the rack. Accordingly, modular power systems may be configured in parallel or in series or a combination thereof to provide multiple power distribution paths of one or more than one. The specific path to the rack may be direct or indirect, and may often depend on the components involved, the amount and type of power, and/or the desired configuration. If desired, the path to the rack may include distributing power to chassis located within the rack. The distributed power may be delivered at one or more desired voltages, which may vary depending on configuration and components. In some cases, the desired voltage may include, for example, 12V, 5V, and/or 3.5V. In some embodiments, when a chassis is used, it may utilize one or more subsystems. Such subsystems may include any desired subsystems that do not interfere with the desired amount and type of power desired to be delivered to the rack. However, as an example, a power-on-package subsystem may be used. Such a package can receive and convert AC current to DC current, and/or vice versa, depending on what is desired. For example, a particularly useful power-on-package subsystem may be configured to accept input power of 208, 240, 380, 400, 415, 480, and/or 600 volts AC and convert it to DC power, for example 48V DC. It can be designed to convert directly.

The modular power supply system or systems may be powered directly or indirectly in any convenient manner. For example, the modular power supply system may be powered directly through a primary electrical distribution system within the chassis. Depending on the type and amount of power and other components, the chassis can provide electrical continuity between the power distribution path and the chassis itself using an interface, for example a set of spring loaded pins or other suitable connector interface. can be built Continuity can then be established between that interface connector and any desired power input interface on a desired server or other computing component located in the chassis. In some embodiments, a power-on-package module may be used within each chassis to convert the voltage to the appropriate level directly in the chassis itself. It can be used for various types of power distribution, but can be particularly useful, for example, in a 48V distribution. 17 shows an exemplary embodiment of a rack power distribution system 950 . In this exemplary embodiment, the rack 310 may receive an AC input 960 at the AC interface 311 of the rack 310 . The power distribution system 950 may generate a DC output 320 and distribute the DC output 320 to one or more chassis 400 .

In some embodiments, ensuring reliable power supply to computer components within the rack is a primary concern. To this end, some embodiments utilize either a blade level power supply or a computer component level power supply, which can supply a specific input voltage and provide the required output voltage to the blade and/or component level power supply. Some embodiments incorporate multiple power sources into each blade to provide redundancy.

In some embodiments, one or more switches may require power. An exemplary switch may be a standard data center grade switch with appropriate interfaces for connecting to the backplane and for providing rack level communications to each blade. Some embodiments distribute only a single voltage, which acts as an interface between the power rails and each blade, directing the voltage to the power input rails or intermediate connectors to be placed between the power power leads and the rack level voltage distribution system. This can be achieved by an interface system and power rails with connectors for delivery via

In some embodiments, there may be one or more power rails that distribute the primary voltage along the bottom of the rack. These rails may be supplied from one or more primary power rectifiers, may similarly be located outside the pressure control vessel, and may be delivered to each rack via cables or busbar systems. At this level, using a higher voltage, for example 48 volts, may reduce the required current carrying capacity of the distribution system and may effectively interface between the distribution rail(s) and the load interface.

In some embodiments, there will be two primary power distribution systems located within the upper structure platform. The first is the Primary Equipment Power System (PEPS) and the second is the Secondary Equipment Power System (SEPS). The purpose of PEPS is to provide electrical service to the components within the container. Such a system may be a high voltage, high current distribution system that accepts an input through a copper conductor or busbar system and delivers it to a primary power source, wherein the primary power source distributes operating current to the chassis, computer components, and/or other critical loads. Responsible for providing equipment. Power will enter the upper structure at a prescribed point, and will terminate with a master service isolation breaker. Upstream of this point will be the electrical service and all power redundancy components used within the system. This input would be a high voltage, such as 415 or 480 volts AC for example. The primary equipment load will be driven by a power source or rectifier, supplied from a breaker panel downstream of the master isolation breaker.

The purpose of the SEPS is to provide electrical service to all of the infrastructure support systems and components located within the upper structure. As components required as part of the secondary equipment infrastructure can expect low input voltages, the SEPS is powered by a step-down transformer, which is connected upstream of the PEPS master service isolation breaker via secondary service isolation. can be supplied.

This arrangement would allow the upper structure support and infrastructure system, including all corresponding components powered from the SEPS, to be turned on and operated without delivering primary power to the rest of the system components. All aspects of the steam control system, as well as the management and control system, may be operated independently of the operation of the PEPS.

In some embodiments, an uninterruptible power supply (UPS) is included as part of or in addition to a power distribution system. The inclusion of a UPS allows the disclosed computing system to operate continuously, even in the event of a temporary interruption to external power.

Components of the disclosed power distribution system may include, but are not limited to, commercially available components such as, for example, uninterruptible power sources, DC power systems, AC power systems, and/or power control and monitoring systems. have. Some such components include, but are not limited to, Vertiv products such as, for example, Liebert and/or Chloride UPS products, double conversion online UPSs, line-interactive UPSs, standby UPSs, lithium-ion battery UPSs, and combinations thereof. can do. UPS products can be single-phase or three-phase. Other exemplary power distribution system components include, for example, Emerson Network Power products, NetSure DC power systems, Vertiv, Liebert, Chloride, and/or NetSure power distribution units, and, for example, inverters, rectifiers, transfer switches, and It may include related components such as combinations thereof. Commercially available monitoring units, controller units, and/or software associated with such components may also be incorporated into certain disclosed embodiments.

Pressure Control Vessels and Pressure Management Systems

An embodiment of the disclosed system includes a pressure control vessel designed to house a two-phase liquid immersion cooling system. The pressure control vessel 110 holds a bath 142 of a dielectric cooling fluid 140 , a condenser 130 with a cooling coil 132 for condensing the gaseous dielectric fluid into a liquid, and a computer component 170 . and house the physical mechanisms and/or equipment necessary to distribute electrical power from the electrical power system to the equipment and components located within the pressure control vessel 110 .

During operation, the pressure control vessel 110 may be maintained at a slight vacuum. It will be appreciated that various special connections and considerations must be made in order to operate a computing system within a pressure controlled vessel 110 maintained at a negative pressure.

Some embodiments of the disclosed system provide a series of fiber optic Media Transfer Protocol (MTP) systems that allow connectivity of fibers into pressure control vessel 110 in addition to exiting panels and cable trays to distribute fibers to racks 310 . use the interface. This arrangement reduces the overall number of penetrations into the pressure control vessel 110 , thereby lowering the likelihood of leaks within the vessel.

Some embodiments of pressure control vessel 110 include sensors to ensure safe operation. Such sensors may include, but are not limited to, a temperature sensor, a fluid level sensor, a pressure sensor 180 , a gas partial pressure sensor, a position sensor, an electrical sensor, a microphone, and/or to automate and/or ensure operation of the system. It may include a camera.

In one exemplary embodiment, the temperature sensor includes, but is not limited to, a sensor for measuring the temperature of a gas phase within the pressure control vessel 110 , a sensor for measuring the temperature of a liquid phase within the pressure control vessel, water and/or other It may include a sensor for measuring the temperature of the process fluid, and/or a sensor for measuring the temperature of other components including the computer component 170 . In some embodiments, thermocouples, thermistors, and/or silicon sensors may be used to measure the temperature of computer components. In some embodiments, the system uses a commonly accepted communication protocol provided by the component itself and such as a device-provided API or other programmatic interface such as JSON over HTTPT or SNMP to determine the equipment temperature, in some embodiments. It may rely on information being retrieved or monitored.

Some embodiments may include various life safety features to ensure user safety. Such features may include, but are not limited to, automatic electromagnetic locking mechanisms, fail safe systems, fire and/or smoke detection and/or suppression systems, ventilation systems, and/or backup lighting. In certain embodiments, these features may be included as part of a generic platform.

Certain embodiments include automatic vapor detection based leak detection systems to ensure that any fluid loss within the pressure controlled vessel is quickly detected. Such a system provides a pressure sensor 180 in the pressure control vessel 110 that monitors the pressure to ensure there is no substantial leak, and/or a pressure that detects the presence of any dielectric vapors that may leak out of the pressure control vessel. It may include a gas sensor disposed outside the control vessel.

The specific design, arrangement, and/or layout of embodiments of the disclosed systems may be adjusted based on the conditions under which such systems are deployed. In some embodiments, the size, materials, internal systems, component mounting and configuration options, and interfaces therebetween of the pressure control vessel 110 , computer components 170 , and power system all dictate the conditions under which the system is used. can be adjusted based on

rack system

8A-8C show an exemplary embodiment of a rack system 310 (or rack 310 ). The rack 310 may serve as an intermediary between the electrical and communication system installed in the pressure control vessel 110 and the computing equipment 170 installed in the rack 310 . To control the spacing, orientation, location, and/or configuration of computer components 170 within pressure control vessel 110 , computer component 170 may be mounted to rack 310 . In one exemplary embodiment, each computer component 170 may be installed within the chassis 400 prior to being mounted on the pressure control vessel 110 .

The rack 310 can be any physical structure that can be used to mount the computer components 170 including, but not limited to, any frame, bracket, support, or other structure. Computer components 170 will be considered mounted to rack 310 when they are attached directly or indirectly to rack 310 and held in a substantially stationary position. Some embodiments use dedicated mechanical guide plates as mounting mechanisms, wiring harnesses attached to bulkhead fittings, and/or through the use of intermediary power and backplane receivers 331 to distribute power and signals within the rack. may include using

The specific design of the rack system 310 may be adjusted based on the conditions under which the system is deployed. Some embodiments of rack 310 may include dedicated switches. In some embodiments, the uplink interfaces may be connected to the computing equipment 170 within the rack, via the fiber infrastructure, and/or the downlink access interfaces via the backplane receiver 331 interface or any other suitable computing equipment connection scheme. can

In certain embodiments, the rack system 310 may include a housing for one or more intermediate power sources capable of distributing an appropriate voltage from the power interface to other equipment installed within the rack 310 . An interface for interconnecting power from a distribution system to an intermediary power source may be incorporated into the design of the rack 310 and thus eliminated by isolating the interface between the various racks, power, and communication systems; and/or Alternatively, an alternative rack configuration may be substituted.

8A shows a top view of the rack 310 . In this exemplary embodiment, the rack 310 includes an AC interface 311 and a data interface 312 . The rack 310 also includes a pair of power sources, a power source 313 and a redundant power source 314 (or backup power source). Rack 310 may also include rectifiers and controllers. Redundant power source 314 (and/or rectifiers and controllers) allows rack 310 to be repaired quickly or continue to function even if the primary power source fails. The rack 310 may optionally include a transducer 315 . The rack 310 is configured to receive the plurality of chassis 400 and to hold the chassis 400 in a substantially stationary position.

In some embodiments, the entire rack 310 may be submerged in a dielectric fluid. This may include submerging the rectifier, power connection, and/or data connection in a dielectric liquid during operation. To reduce and/or prevent plastic contamination of the dielectric fluid, in some embodiments, plastic insulation and/or cable shielding may be removed. In some embodiments, the dielectric fluid may serve to insulate otherwise exposed cables and/or connections.

8B shows a perspective view of a rack 310 including a plurality of chassis 400 . The disclosed configuration of the rack promotes hot swappability of the chassis 400 . In this exemplary embodiment, rack 310 may include a plurality of AC cables 318 that connect AC interface 311 to power source 313 and/or redundant power source 314 . Power supply 313 and/or redundant power supply 314 may generate DC output 320 , which may be delivered to backplane receiver 331 via DC cable 321 . Rack 310 may also include a plurality of data cables 319 connecting data interface 312 to backplane receiver 331 . The backplane receiver 331 may be used to supply data from a data connection on the lower end of the chassis 400 to a data connection located at the upper end of the rack.

8C shows a side view of the rack 310 . In some embodiments, rack 310 provides mechanical stabilization and/or housing for chassis 400 and its components. The rack 310 also provides routing of power and data cables from an upper end of the rack 310 that is generally accessible within the container to a lower end of the rack 310 where the cables are connected with the chassis 400 . helps

Chassis and interface system

In one exemplary embodiment, the purpose of the disclosed chassis system 400 is to serve as a standardized physical intermediary component between the conventional and/or intentionally constructed computing components 170 and the disclosed rack system 310 . will do In one exemplary embodiment, the purpose of the backplane receiver 331 is to provide a slot-in interface between the chassis 400 and the rack 310 , to provide a power source within the power system and various computing components installed within the chassis 400 . 170 to allow power and signal distribution between network switches in the communication system.

In some embodiments, the pressure control vessel of the present disclosure may include at least one rack 310 that may contain one or more servers, eg, blade servers. Each server may be attached to a chassis 400 (also referred to as a server case or case). 9A-9G show an exemplary embodiment of a chassis 400 for mounting various components 170 . The chassis may facilitate the installation of the server on a rack of pressure control vessels or its removal from the system. In some embodiments, other electronic components of the pressure control vessel may be mounted within the chassis. For example, any part of a computer component or hardware, such as a motherboard, chip, card, GPU or CPU, may be installed within the chassis. As another example, components such as power sources, power interfaces, or network communication interfaces may be mounted within the chassis.

In one exemplary embodiment, the chassis may serve as a common interface between a component (eg, a server) and a pressure control vessel. The chassis can provide a variety of mounting, power, and connectivity features that can be customized based on the characteristics or design of the components. In other words, various aspects of the chassis may be modified based on the design specifications of the components. Thus, the chassis can accommodate almost any model or type of hardware. For example, a chassis may facilitate the use of specially designed hardware or off-the-shelf hardware.

Embodiments of the chassis 400 may include components designed to allow for the application of existing commercial components, the use of customer designed components, and/or the use of a particular chassis for a particular application. Embodiments may include standard motherboards and adaptation kits for special components. In certain embodiments, such components include Gigabyte motherboards with NVidia GPUs and/or Supermicro motherboards with Intel CPUs.

9A illustrates a chassis 400 for rack mounting a server in accordance with an exemplary embodiment. In this exemplary embodiment, the chassis 400 may be a rectangular box including a rear wall 410 and two side walls 420 . The rear wall 410 may include a plurality of holes 411 for facilitating the circulation of a fluid in the chassis 400 . The chassis 400 may include guide rails 421 on each sidewall 420 .

9B illustrates a plurality of components inside the chassis 400 , according to an exemplary embodiment. In this exemplary embodiment, the rear wall 410 is removed. Accordingly, FIG. 9B shows a server 430 , including a power supply module 431 , a GPU module 432 , a CPU module 433 , and an interface card 434 . In one exemplary embodiment, the components in the chassis 400 may include components used in the blade server, for example, a CPU module 433 and a GPU module 432 . In addition, the components inside the chassis 400 may include other components not normally included in the server, for example, the power module 431 or the interface card 434 . Because the chassis 400 does not require conventional air cooling equipment, the chassis 400 does not include any fans or heat sinks within the chassis. Thus, the chassis has a very thin profile compared to the computing power of the chassis.

9c shows a schematic diagram of the components within the chassis. In this exemplary embodiment, a server motherboard 445 , a plurality of power modules 431 , and an interface card 434 are mounted on the chassis 400 . Storage devices and/or other peripheral components may also be mounted to chassis 400 , along with backplane interface 330 and/or power modules and communication system modules.

In one example, a mounting interface may be added to or removed from the chassis such that a piece of hardware is secured to the chassis. On the interior surface of the chassis 400 , provision may be made to allow components (eg, motherboards, GPUs, CPUs, interface cards, and other related components) to be mounted to the chassis. This provision is a mounting interface. The specific arrangement of the chassis system 400 may vary depending on the equipment and/or components to be attached to the chassis 400 and/or racks. Some embodiments of chassis 400 may feature interchangeable mounting plates that may be used for equipment attachment. A set of standard attachment plates are available for common or frequently used components.

The style or form factor of the power and network interface modules within the chassis system 400 may be adjusted based on needs and requirements regarding specific components and/or user specific equipment. In one example, the power subsystem of the chassis may be modified to address the need for a particular component. In another example, the size of the chassis may be designed to accommodate a piece of hardware of any size. In another example, the chassis may provide different networking options, depending on the network connection card installed in the chassis. Due to these and other features of the chassis, the chassis can accommodate a variety of components. As a result, the assembly or removal of these components of the pressure control vessel can be simplified and thus can be automated. For example, a chassis may include a blade server, and a robot may easily install or remove the chassis from a rack of a pressure control vessel. Thus, the robot can remove and replace the blade server without any human intervention, thereby minimizing human exposure to the dielectric fluid.

In one exemplary embodiment, the chassis may include a microcontroller capable of communicating with a management system of the pressure control vessel. The microcontroller may receive sensor data from various sensors disposed inside or outside the chassis. For example, the chassis may include a sensor to detect whether the chassis has been properly positioned within the rack. In the case where the server can be connected to the rack, the server is appropriately placed within the rack. The sensor may determine whether the chassis is properly positioned within the rack. Thus, the sensor can send data to the microcontroller, and using the data, the microcontroller can provide a signal to the management system that indicates whether the chassis has been properly positioned within the rack.

In one embodiment, the microcontroller may be coupled to a switch that may turn on or off power to a component mounted within the chassis. The microcontroller may receive a power on or off signal from the management system, and in response to receiving such signal, the microcontroller may send a signal to the switch to power on or off a component, eg, a server. In one exemplary embodiment, the microcontroller may receive operational data from a server, and the microcontroller may relay such data to the management system. Operational data is a key performance indicator of a server, and may indicate its performance. Operational data may include the speed of the computer, the degradation of the computer, power consumption, the temperature of the circuit, and the bandwidth of the system.

In one exemplary embodiment, the microcontroller may monitor, manage, and control the electrical and communication facilities of the blade server. For example, by monitoring indicators such as electrical current (ie, amperage) and voltage, it is ensured that the system can protect itself and, for example, that over-current or under-current is not provided.

In one exemplary embodiment, the chassis may include structures that allow the robot to grip and remove the chassis. For example, the chassis may be in the shape of a rectangular box having a front portion, a rear portion and sidewalls. The chassis may also include a top wall and a bottom wall. The top wall of the chassis may include a plate that may be coupled with the robot arm. When using the plate, the robotic arm can grip the plate for unloading and other handling operations.

In one exemplary embodiment, the chassis may include mechanical guide rails and placement pins to ensure proper alignment and insertion of the chassis within the rack. A mechanical guide rail may be arranged on a side wall of the chassis.

In one exemplary embodiment, the chassis may include various features to facilitate fluid flow. For example, the chassis may be in the shape of a rectangular box having a front portion, a rear portion and sidewalls. The chassis may also include a top wall and a bottom wall. In this example, at least one of the walls of the chassis may include a fluid flow hole through the entire wall. For example, the rear wall can include a plurality of holes that can facilitate fluid flow in and out of the chassis when the chassis is immersed in a liquid bath.

In one exemplary embodiment, the chassis may include an opening to ensure drainage of all fluids within the chassis when the chassis is removed from the liquid bath. For example, the rack may be placed in a liquid bath to cool the computer components held by the rack. To remove the server, the robot may grip the plate of the chassis and lift the chassis out of the rack (thereby removing the chassis from the liquid bath). When the chassis is removed from the liquid bath, a certain amount of fluid may remain within the chassis. The chassis may include a notch or drain in the bottom wall of the chassis to ensure that fluid can escape even if the pressure control vessel is not perfectly flat. The notch or drain may be located at the edge of the bottom wall.

In one exemplary embodiment, the chassis may include a power interface and/or a communication interface. Such interfaces may electrically couple components mounted within the chassis to racks and/or pressure control vessels. The power interface and/or communication interface may be disposed on the backplane. For example, a server mounted within a chassis may be connected to an interface of the chassis via various wires and cables. When the chassis is placed within the rack, the interface may be electrically coupled to the rack (ie, the backplane receiver) and/or other interfaces connected to the pressure control vessel. Electrical coupling between the two interfaces (eg, a backplane and a backplane receiver) may provide power to the server and connect the server to a communication network within or outside the pressure control vessel. The coupling between the two interfaces can take place automatically during mechanical insertion of the chassis in the rack. Similarly, removal of the chassis from the rack may separate the server from the rack and/or pressure control vessel.

In some embodiments, providing standardized interconnectivity through the backplane interface 330 and the communication system interface may minimize the chance of data interfaces being misconnected and may reduce the need to troubleshoot connectivity issues.

In certain embodiments, chassis 400 will include a set of standard power and network interfaces. The network interface may be in the form of a Cat6A or Cat7 compatible RJ45 interface for connection to a 1G or 10G Ethernet interface on the equipment motherboard. In such embodiments, the power interface may include a set of standard Molex style connectors for connection of standard motherboard and/or peripheral components.

In one exemplary embodiment, the pressure control vessel may include an internal database that stores information regarding components installed in the system. The internal database may be a repository of components installed in the pressure control vessel. For example, an internal database may store the make and model of all servers and power supplies installed in the system. Since components of the system can be exchanged or replaced, for example, by a robot, the management system can keep track of changes and update information stored in an internal database. The pressure control vessel may be connected to an external database via a network.

In one exemplary embodiment, each chassis may be associated with a unique serial number, eg displayed as a barcode on the chassis. When a component is placed within the chassis, the component's specification (or component's make and model) may be associated with a unique serial number and stored in an external database. Thereafter, when the chassis is installed in the pressure control vessel, the pressure control vessel can retrieve the component by looking up the unique serial number in an external database. For example, the robotic arm could scan a barcode on the chassis, and the management system could use the barcode to search an external database. The management system may update the internal database using information obtained from the external database. Similarly, when the chassis is removed from the pressure control vessel, the robotic arm can scan a barcode associated with the chassis, and the management system updates its internal database to indicate that a component mounted on the chassis is no longer installed within the system. can indicate

In one exemplary embodiment, the chassis may include an RFID tag. The robotic arm of the pressure controlled vessel may include a scanner capable of emitting radio frequencies to detect the RFID tag. When the robot arm handles the chassis, it can scan the RDIF tag and provide a unique serial number to the management system to update its internal database.

In one exemplary embodiment, the chassis may include an identification plate that may include a user specific asset identification number. This asset identification number may be stored in association with a component mounted within the chassis. In some embodiments, the identification plate may be a chip configured to store an asset identification number.

In one exemplary embodiment, the chassis may include a pump to enhance the flow of fluid within the chassis. To maximize heat exchange between the liquid bath and components within the chassis, the chassis may include a pump capable of circulating fluid within and around the components. The pump may draw fluid from several conduits that run around the chassis and may propel the fluid out of the chassis or vice versa.

In one exemplary embodiment, the chassis may include several conduits around the chassis for drying the chassis and components mounted therein. As the chassis is pulled from the liquid bath, an amount of liquid may remain within the chassis or components therein. The chassis may include several conduits that may guide the flow of gases within or around the chassis to facilitate drying of the chassis and components. In one exemplary embodiment, the pressure control vessel may expose the chassis to a flow of gas prior to delivering the chassis to a user. For example, the chassis may include an input pipe for receiving a flow of gas, and the pressure control vessel may provide a flow of gas through the input pipe.

9D shows the bottom wall 415 of the chassis 400 , according to an exemplary embodiment. In this exemplary embodiment, the bottom wall 415 may include a power interface 416 and a communication interface 417 . 9D shows the guide rails 421 on the sidewalls 420 of the chassis 400 .

9E shows a top wall 425 of the chassis 400 , according to an exemplary embodiment. In this exemplary embodiment, the top wall 425 may include a pair of plates 426 and handles 427 . The robot arm may use the plate 426 to lift the chassis 400 .

9F shows a sidewall 420 of the chassis 400 , according to an exemplary embodiment. In this exemplary embodiment, the sidewall 420 may include a guide rail 421 . 9F also shows the rear wall 410 , the handle 427 , and the power interface 416 .

9G shows an exploded view of the lower drain hole 450 of the chassis 400 , according to an exemplary embodiment. In this exemplary embodiment, the bottom drain hole 450 may be disposed at the corners of the bottom wall 415 , the side wall 420 , and the rear wall 410 .

10A-10F show an exemplary embodiment of a pressure control vessel 500 . In particular, FIG. 10A shows an exemplary embodiment of a vessel 500 , for example a 600 KW skid. Exemplary embodiments include a modular skid. The container 500 may include a plurality of tubes 514 for forklifts, which aid in movement and delivery of the container 500 to a desired location. Vessel 500 may receive process water from power and communication inputs 511 and process water pipe 512 with minimal penetration through the vessel itself. Such connections may be placed at the top of the vessel, thus facilitating close packing of the modular vessel within the data center. In some embodiments, to accommodate vertical stacking of multiple modular containers within a data center, connections may be placed on the front and/or sides of the containers. In some embodiments, the containers may include vertical spacing to help the containers stack vertically on top of each other. The vertical space can create additional space for connections, air flow, and/or insulation between containers. By vertically stacking containers, good power density can be achieved on a square foot basis. In some embodiments, vessel 500 may include a power and communication box configured to receive input 511 and distributed power and network connectivity throughout vessel 500 . The container 500 may include a sealing lid 515 , which may facilitate the addition and/or removal of components to the container 500 .

10B shows another view of the container 500 . In some embodiments, an inventory of replacement components may be stored in the container 500 such that the components may be replaced using a robotic system within the container without opening the container. The robotic system may be operated using a gantry motor 516 . In such embodiments, when a component is destroyed or in need of repair, a replacement component is installed into the system and the destroyed or otherwise removed component can be stored in the cassette until the cassette is completely filled. At the point of full filling, the cassette containing the removed components can be removed from the container and a new cassette with new replacement components can be inserted into the container for later use. In some embodiments, the disclosed container has a length of about 15 feet, a width of about 7 feet, and a height of about 10 feet. In some embodiments, the disclosed system may provide for the achievement of 600 KW of computing power within about 150 square feet.

In some embodiments, vessel 500 may also include one or more bellows tanks 517 . A bellows tank 517 may be used to regulate the pressure within the vessel. When the disclosed computing and/or cooling system is initially activated, the expanding dielectric fluid may be directed to the bellows tank, thus not being lost to the environment and/or preventing pressure from accumulating within the vessel. In some embodiments, bellows tank 517 may be large enough to hold about twice the amount of liquid dielectric fluid in the container.

10C shows a cross-sectional view of vessel 500 . The lower portion of the container 500 may include a rack 310 and/or a chassis 400 that houses computing components. Above the rack is a condenser coil 132 that cools and condenses any dielectric vapor. Power may be distributed within the vessel using a power bus bar 518 . This allows power to be distributed to the individual computing components in a highly-exchangeable manner. Power bus bar 518 allows the vessel to receive external power using only one or a small number of penetrations through the vessel. This design simplifies the installation and operation of the vessel system. In some embodiments, each power bus bar can provide 600 amps for power to 5 racks. In such an embodiment, there may be two sets of bus bars, one on each side of the container. In some embodiments, the bus bar does not include plastic insulation. Plastics may be considered a contamination of some dielectric fluids, and in some embodiments may be generally avoided.

In some embodiments, container 500 may include desiccant 519 . In some embodiments, the dielectric vapor may be removed from the headspace of the vessel 500 and condensed in a manner that allows any non-condensable components to be removed from the dielectric fluid. Water will not condense under the same conditions as many dielectric fluids. Accordingly, such a system can be used to remove water contaminants from dielectric fluids.

In some embodiments, vessel 500 may include a fluid filter 520 , a fluid pipe 521 , and a fluid pump 522 . In some embodiments, the dielectric fluid may be added to the vessel in a manner that causes the liquid dielectric fluid to flow out of the rack 310 and into the sump area 523 . The fluid may then be filtered using fluid filter 520 and pumped to the distal side of the vessel using fluid pump 522 and fluid pipe 521 . Such a system circulates fresh, filtered dielectric fluid through the vessel, so that the dielectric fluid can be reused to cool computing components.

10D shows a cross-sectional view of vessel 500 . In such an embodiment, the level of the liquid dielectric fluid may be maintained at a fluid level 524 that is higher than the height of the rack 310 and/or computing components therein. As a result, the rack 310 and/or computing components may be submerged in the dielectric fluid. Above fluid level 524 , for example up to half level 525 , there may be saturated dielectric vapor. In some embodiments, the saturated dielectric vapor is maintained below half level 525 , which may be about half the height of condenser coil 132 . Above the saturated vapor is located a headspace that, in some embodiments, may contain a less dense dielectric vapor.

In the exemplary embodiment of FIG. 10D , cooling coils 132 are located above the shelf area. Thus, when the robot 526 places or removes the chassis 400 , the cooling coils 132 are not positioned in the path. This arrangement of cooling coils 132 may streamline the placement and removal of chassis 400 , thereby providing significant advantages in autonomous operation of the vessel.

communication system

Embodiments of the disclosed communication system are designed to provide standardized Layer 1 to 3 connectivity and management interfaces for equipment within or associated with the disclosed superstructure 210 , pressure control vessel 110 , and/or computing system. .

In some embodiments, a series of MTP interfaces provides the ability to bring multiple high-density multimodal fiber connections into the pressure control vessel 110 . Once housed within the pressure control vessel 110 , the fiber connections are made to individual switch level connections, using a set of dedicated breakout cables, a breakout interface, a patch panel, and/or a distribution patch panel to the rack 310 . can be divided into

Some embodiments of the disclosed system may include a dedicated fiber patch panel interface port in each rack 310, thus allowing connection to an internally installed switch system via a short patch panel. In other embodiments, there may be a dedicated patch panel, or set of patch panels, extending from each switch system to the MTP distribution interface.

In some embodiments, the interface between the switch system and chassis 400 may be through a backplane interface 330 and/or through some other mechanism that may or may not include the use of a backplane connector. In some embodiments, there may be no intervening rack level switch system. Such an embodiment may utilize a centralized set of switches within the pressure control vessel 110 to connect to various computing equipment located therein.

A typical interface between the switch system and chassis 400 is using a patch panel that is attached to a rack 310 and wired to the backplane system 330 with patch cables that connect ports on the patch panels to appropriate ports on the switch system. can be achieved.

In some embodiments, patch panels interconnecting the communications system cabinet with the MTP interface on each pressure control vessel 110, and a centralized communications system distribution serving to interconnect the switch systems to each other and/or to the outside world. There will be a small (6U) rack rail area, containing the switches. In such embodiments, end-users or customers install their own routing gear within this space and serve as a link between the disclosed computing system and the outside world, or pressure control vessel 110 or upper structure 210 . ) and the existing network environment may choose to provide them with external connectivity to enable fiber connectivity.

The access, communication, and/or networking components used in embodiments of the communication system environment may be standard equipment or may be user-specific. The rack 310 and backplane interface 330 system involves removing the existing switch, replacing it with any standard switch (eg, a 1U switch), and re-wiring the desired interface to the backplane network interface panel. By connecting, this may include the ability to replace a switch system located within each rack 310 .

In certain embodiments, products designed to interface directly with the backplane system 330 may be used. Such a product is a chassis 400, specifically designed to interconnect switch ports, either through a purpose-built internetworking interface, through a commercially available protocol, or through a specification for the design of a network level interconnectivity interface. Patch panel systems and/or direct electrical interfaces may be used.

In some embodiments, the connectivity between each blade or chassis and the switch may include multiple interfaces. One interface may be a standard switchport, which may be a standard port available on a commercially available switch. The common interface may be 1GBASE-T or 10GBASE-T, allowing the use of Cat6 or Cat7 twisted pair copper connections between the switch and the host device. Another interface may consist of a patch panel with standard patch cables running from a standard switchport to the front side of the patch panel and a set of hard-wired connections from the back side of that patch panel to the signal interface of the signal backplane. It may be a switch-to-backplane intermediary device. Alternatively, it may consist of a special cable and/or standard RJ45 interface running from the switchport to the board to establish connectivity between the standard switchport and the backplane. Another interface may be an interface system signal backplane that distributes signal paths from standard switchports along the printed circuit board (PCB). One or more of the signal paths may terminate at a connector on the PCB to which the signal backplane interface is to be connected. Another interface may be a chassis signal backplane interface. This may be a connector located on the chassis itself that mates with a connector on the interface system signal backplane. It serves as an interface between the interface system signal backplane and the chassis itself. Another interface may be a chassis network interface. This can be a standard patch interface that allows the connection of patch cables from the chassis network interface to the RJ45 interface on the server attached to the chassis.

robot system

In some embodiments of the disclosed system, a potential method of addressing the need for good interchangeability of components within the pressure control vessel 110 . The need for the ability to remotely remove and replace a failed component 170 may be addressed via a robot.

Certain embodiments of combinations of the disclosed systems may include an inner robotic arm 230 and/or an outer robotic arm 240 . In some embodiments, for example, for cryptocurrency applications and/or specific high-performance computing environments may not require good exchangeability of components. In other hyperscale GPU and CPU environments, this may be a basic requirement. Embodiments of the disclosed robotic system allow for replacement of chassis and/or other computer components without interruption of any other components. In some embodiments, failed cards and/or components may be automatically and/or programmatically replaced and/or stored. This allows for short and mid-duration, fully spaced-apart and autonomous operation of embodiments of the disclosed system.

The inner robotic arm 230 mechanism is located within the pressure control vessel 110 environment. 7A-7D , in an exemplary embodiment, when a card or component is not operating properly, a removal sequence may be initiated. When the removal sequence is initiated, the inner arm 230 will remove the appropriate computer component 170 and/or associated chassis 400 from the rack 310 , which airlock positioned within the pressure control vessel 110 . will move to 220 , which will signal the completion of the removal sequence. Once this sequence is complete, the inner airlock door 222 will close, the airlock pressure will equal the pressure of the outer atmosphere, and the outer airlock door 224 will open. Once the outer door 224 is opened, the outer robotic arm 240 will remove the chassis 400 from the airlock 220 and place it into an empty storage slot.

In some embodiments, before the airlock 220 is opened to the outside environment, the airlock 220 will be purged with nitrogen and/or other inert and/or non-condensable gas. In such embodiments, this has the effect of reducing or preventing the loss of dielectric vapor when the airlock is opened and closed. In certain embodiments, the airlock will be fitted with a one-way valve, either on the inner portion, the outer portion, or both. In embodiments having one-way valves on both the inner and outer portions of the airlock, the purge of the airlock will prevent cross-contamination of the external environment into the atmosphere inside the pressure control vessel 110, and will also prevent loss of dielectric vapors. will be.

When the card or component replacement sequence is initiated, the external robotic arm 240 will remove the replacement component and/or chassis 400 from the storage slot and place the component within the airlock 220 . Once complete, the outer airlock door 224 will close, the airlock pressure will equal the pressure inside the pressure control vessel 110 , and the inner door 222 will open. Once the inner door 222 is opened, the inner robotic arm 230 will remove the chassis 400 from the airlock 220 and insert it into a suitable rack 310 .

When coupled with a remotely accessible management system, the inner and outer robotic arms 230 and 240 enable remote operation and management of the data center environment. This may reduce the need for a human operator to remain available, and may reduce cost and/or downtime. In some embodiments, external robotic arm 240 is mounted to a movable base, thereby allowing a single external robotic arm system to serve for multiple embodiments of the disclosed computing system.

When integrated with customer-developed workflow management systems and virtualization technologies, the disclosed robotic systems enable the development of fully autonomous, self-healing data center solutions, which can provide the highest level of system reliability.

In some embodiments, an asset tag and/or production batch code with a unique human and/or machine readable serial number may be included in each computer component and/or chassis. In such embodiments, the asset tag may be a unique serial number. The tag may include a printed barcode or QR code and may enable automatic part identification by embodiments of the disclosed robotic system. Tag codes can also be used with management software systems to provide detailed component information related to inventory management and automation systems. The tag and any associated adhesive or other components are preferably made of a material compatible with the dielectric fluid. The tag is preferably positioned on the chassis in a spot that can be read when the chassis is inserted into the rack. In some embodiments, secondary or additional tags may be located on other areas of the chassis to aid in component identification and/or inventory management.

Embodiments of the disclosed robotic system allow any individual chassis to be temporarily removed and replaced, a process referred to as “re-seating”. This is useful when, during troubleshooting, it is determined that a component's hard power cycle is required. Reseat accomplishes this by disconnecting all power, waiting a bit, and reconnecting it.

Some embodiments allow individual cards and/or chassis to be removed from the pressure control vessel via an airlock. In some embodiments, the robotic system will remove the chassis from its slot in the rack, move it to the airlock, and complete these tasks so that the airlock can be opened and the card and/or chassis can be removed. will signal Some embodiments allow replacement components and/or chassis to be placed into specific rack slots, via the same airlocks used for removal. In some embodiments, the robotic system will remove the chassis from the airlock, place it in the appropriate rack slot, and signal completion of this task.

Robot on the inner system

Embodiments of the disclosed system may include "inside the robot" robotic system. In such an embodiment, the pressure control vessel may be expandable to accommodate a robotic arm operating within the vessel. The container may also be arranged to accommodate movement or transfer of computer components and/or chassis over racks that contain the computer components being operated on. It will be understood that a pressure control vessel may also be referred to as a tank, a pod, and/or a vacuum chamber. Alternatively, it will be understood that certain components of the pressure control vessel may be referred to as tanks or pods.

10E shows an embodiment of the disclosed system with a gantry robot 526 configured to remove, replace, and/or install a computing component, eg, a chassis 400 of a rack 310 . In some embodiments, gantry robot 526 may also be configured to remove, replace, and/or install DC rectifiers and/or other components of the power distribution system. It will be appreciated that some embodiments of the disclosed computer components and power distribution components may be designed to be highly interchangeable and may include handles or other features to facilitate handling by the gantry robot 526 . In some embodiments, the gantry robot 526 is arranged to move in both the x and y directions and may be dropped in the z direction to remove and/or install replacement components. In some embodiments, the gantry robot 526 includes a gripping tool for gripping the chassis 400 and/or the power source, eg, the gripping tool may grip the plate 426 .

10E shows a top cross-sectional view of an exemplary embodiment of the disclosed tank. In some embodiments, an array of racks 310 may be populated with chassis 400 and/or computing boards. In some embodiments, each chassis 400 may utilize about 6KW of power, and each rack 310 may contain 10 chassis. Thus, in an embodiment comprising ten such racks 310 , the container may utilize about 600 KW of power for computing purposes. In some embodiments, additional rack 310 and/or magazine 527 of chassis 400 and DC power rectifier may be stored within vessel 500 for use as replacement components, and/or vessel 500 ) may provide space for storing components removed from

Robot on the outer system

12A-12E show another embodiment of a container. In particular, FIG. 12A shows an embodiment of a container 700 , in which a gantry robot 526 is located outside of a tank 710 that houses the chassis 400 and/or computing components. In this embodiment, the tank 710 may be small, but will need to be opened more frequently to allow the external gantry robot 526 to access the chassis 400 and/or power source inside the tank 710 . Additionally, replacement equipment may be stored and/or housed within a modular enclosure, such as storage 716 on the exterior of tank 710 . In some embodiments, tank 710 may have multiple doors 711 , whereby when a single door 711 is opened for removal, installation, and/or replacement of components or chassis 400 . , it is possible to limit the exposure of the inner portion of the tank (710). In such embodiments, replacement components may be stored outside of tank 710 to prevent unnecessarily opening the tank.

Vessel 700 may also include one or more transformers 712 , electrical distribution panel 713 , process water pipes 512 , and electrical chase 714 . Vessel 700 may also include a programmable logic controller (PLC) cabinet 715 that monitors and controls the status of various equipment within vessel 700 . A transformer 712 , an electrical distribution panel 713 , a process water pipe 512 and an electrical chase 714 and a PLC cabinet 715 may be located outside of the tank 710 .

12B shows a cross-sectional view of a vessel 700 with a tank 710 accessible to an external gantry robot 526 . In this exemplary embodiment, condenser coil 132 , rack 310 and bellows 717 are located within tank 710 . 12C shows a side view of a vessel 700 with an external gantry robot and a tank 710 with multiple doors 711 . In this exemplary embodiment, tank 710 includes a fluid pump for removing fluid from the sump region and directing the fluid through fluid pipe 521 to fluid filter 520 . The container 700 also includes a magazine 718 for storing replacement equipment. In this exemplary embodiment, the magazine 718 is located on the outside of the tank 710 . In some embodiments, spacers and/or ballast blocks 160 may be used to reduce the total volume of liquid dielectric fluid in tank 710 .

12D shows a rack 310 according to an exemplary embodiment. In some embodiments, redundant power sources 314 may be disposed on opposite sides of rack 310 , adjacent and opposite primary power sources 313 . In addition, power and/or data cables 318 and 319 may be routed in alternative configurations, thus accommodating the specific requirements of a particular deployment. In this exemplary embodiment, the backplane receiver 331 is located below the rack 310 .

12E shows an exemplary hinged door 711 that may be used in some alternative embodiments of the disclosed tank 710 . In some embodiments, sliding doors may be used, as opposed to hinged doors, to reduce or prevent flow induction in the dielectric vapor. Slowly sliding the door open will interfere less with the dielectric vapor than rotating the hinged door and mixing the flow.

management system

The management system is a web-based interface between a user of the disclosed computing system and the computing system itself. Embodiments of a management system provide an operational view of a computing system, and include monitoring and management of pressure control vessels 110, robotic systems, communication systems, power systems, and/or all other systems and components. Enables monitoring and management of components. In one exemplary embodiment, the management system may be implemented within the PLC cabinet 715 of FIG. 12A . In another exemplary embodiment, the management system may be implemented within the power and communication box 513 of FIG. 10A . In each embodiment, the power management system may be implemented as a software program on a control device or other suitable device, for example a computer.

In certain embodiments, a simple network management protocol may make a set of accessible data points available to users of the management system, thereby enabling monitoring of key operating parameters through a third party monitoring system. A full operation log may be maintained, and a chart may be provided for review of the user's operation condition data.

Regular maintenance of system components can be planned and maintained through the management system. The user may be provided with regular reminders of routine maintenance and will be aware of such maintenance being performed within the interface. All of these data may be maintained as part of the operation log information for historical operation review.

In some embodiments, operational functionality may also be exposed through API interfaces, thus enabling remote programmatic monitoring and management of computing systems and associated components. A full set of motion monitoring and alerting functions may be included to notify the operator in case of any occurrence.

A centralized server version of the management system or a hosted cloud-based managed version may be used by customers with multiple pressure control vessel computing systems. This provides the operator with one programmatic and user-accessible interface for management of a fleet of pressure control vessel computing systems.

In some embodiments, the software-based interface module enables interoperability with the computing platform and third-party management utilities, such as Microsoft System Center and VMWare VCenter. The user and API interfaces provided by the management system enable full interoperability with the disclosed robotic system, thus enabling fully remote and programmatic autonomous operation and management of the disclosed computing platform.

In some embodiments, the control system enables coordination and control of operations, including temperature, pressure, flow rate, and/or power management. In some embodiments, a user authentication system allows multiple unique users to authenticate to the system. Some embodiments include role-based and/or element-based permission systems. In such embodiments, an administrator may configure multiple roles to which users may be associated and/or may apply specific permissions to individual users outside of the user's role assignments.

Some embodiments incorporate video management to provide users with the ability to record and retrieve video feeds from cameras that may be located within containers and/or upper structures. In some embodiments, the camera may acquire visual data that may be analyzed by the processor. In such embodiments, the processor may use computer vision technology to control the operation of the vessel system, robotic system, and/or upper structure system in response to the obtained visual data.

In some embodiments, the control system and software may be configured to generate reports related to the operation and status of the overall system and/or individual subsystems and/or components of the disclosed computing platform.

Exemplary combined system embodiments

It will be appreciated that the disclosed systems may be used individually or in combination. There are a number of embodiments of a combined computing system that can be tailored to different use cases.

One exemplary embodiment is the Crypto Series. This includes purposefully built computing hardware, a rack 310 with guide plates and wiring harnesses designed for that hardware, a modified implementation of the communication system 360 architecture, and a 1MW pressure control vessel 110 and power distribution. It is an ultra-high density implementation of the disclosed technology using the system. A typical user of this embodiment would be anyone looking to perform cryptocurrency mining or other ultra-high power density processes using custom computing components, or developing a full range two-phase liquid immersion cooling system that would like to integrate their hardware. The manufacturer of the desired computing component.

Another exemplary embodiment is the GPU Series. This is a high-density GPU supercomputing implementation of the disclosed technology. This implementation is a customer-manufactured chassis 400, rack 310 and backplane, designed to integrate a motherboard from Gigabyte and a GPU from NVidia, using NVidia NVLink technology to facilitate ultra-fast GPU-to-GPU communication. interface 330 technology will be available. Typical users of these technologies include general-purpose parallel processing applications that enable GPU-based computing and memory capabilities, including graphics rendering, particle simulation, and general research activities.

Another exemplary embodiment is the CPU Series. This is a high-density CPU computing implementation of the disclosed technology. These implementations will enable high-performance Supermicro-based motherboards, Intel Xeon CPUs, high-speed network interfaces, high-speed memory, and solid-state storage for local storage. Typical users of this technology include data centers, enterprises, and cloud/VPS hosting providers and service providers, who use high-performance computing for their own internal applications or for providing them to third-party customers and other organizations.

Another exemplary embodiment includes the Edge Series. This is a scaled-down implementation of the disclosed computing system, specifically designed for remote/field deployments or within or with typical business and data center environments. This embodiment focuses on a robust weather-resistant environment with full remote monitoring and management capabilities. Target users of these technologies are network operators and other organizations with distributed field infrastructure, and operators of existing facilities who wish to augment their computing power with minimal modifications to existing facilities or structures. It can be a manipulator of technology. Such systems may incorporate various enhancements to external structures to simplify the connection of utility services, including electrical, water and network connectivity to the system.

standalone embodiment

Some disclosed embodiments do not rely on an external water source. Such an embodiment may include a closed-loop chiller for cooling water or other fluid that may be circulated through the condenser as described above. The use of a closed-loop chiller instead of an external source of cooling water enables a substantially independent embodiment.

13 shows an example of a free-standing container 750 . The exemplary embodiment of FIG. 13 utilizes a skid-mounted closed loop chiller 719 to allow coolant or other liquids to be utilized in the condenser within the pod or submersion tank 710 . The use of a closed loop chiller may eliminate the need for an external source of cooling water, thus resulting in a standalone data center solution that may only require an external source of power and a network connection for full operation. Vessel 750 may also include bellows 717 , door 711 , gantry robot 526 , electrical distribution panel 713 , PLC cabinet 715 and magazine 718 .

In some embodiments, closed loop chiller 719 may be a skid-mounted closed loop chiller enclosed within the outer housing of the modular pressure control vessel. In such an embodiment, heat is transferred from the computer components to the dielectric liquid in the tank 710 . This process converts the dielectric liquid into a dielectric vapor as described above. The dielectric vapor rises in tank 710 and is cooled by a condenser, thereby converting the dielectric vapor back to a dielectric liquid. Heat transferred from the dielectric vapor to the condenser is then transferred from the condenser to the refrigerant or condensing fluid in the condenser and then to the closed loop chiller 719 . In some embodiments, chiller 719 removes heat from the refrigerant or condensed fluid using vapor-compression, compressor, evaporator, heat-exchanger, or other closed-loop method of cooling the refrigerant or condensed fluid. Heat from the refrigerant or condensed fluid is finally dissipated through air cooling. In some embodiments, this results in a standalone, modular, air-cooled, two-phase liquid immersion computing system. The air cooling of some stand-alone embodiments is surprising, since the field of immersion cooling is generally known as a contrast to air cooling, particularly air cooling of stand-alone units.

Some disclosed embodiments may be provided in a form factor with a space-saving footprint. An exemplary embodiment includes a single rack comprising ten blades or servers immersed in a dielectric liquid as described above. In some embodiments, each server may consume about 6 kW of power. Accordingly, some example embodiments provide about 60 kW of computer power within a small footprint.

The exemplary embodiment shown in FIG. 13 is housed within a footprint of about 4 feet 2 inches deep, about 8 feet 8.5 inches wide, and about 8 feet 8 inches high. This exemplary embodiment contains about 60 kW of computer power, as well as other operating components and systems, and is housed within an area of about 36.3 square feet. It will be appreciated that the operating components of the vessel may include, but are not limited to, a tank or pod containing dielectric fluid, condenser, power, and data connections for computer components. Such vessels may also contain sensors, control equipment, power cabinets, bellows 717 , vacuum systems, fluid filters, purge systems, and/or other components. Some standalone embodiments may include an outer housing. In some embodiments, the outer housing may enclose the container, may provide structural support, may be skid mounted, may be breathable, may be weatherproof and/or waterproof, and/or may be decorative. have. In some embodiments, the outer housing of the free-standing vessel may include a radiator coil, fan grate, heat transfer component, and/or air-cooled component to facilitate the use of a closed loop chiller. .

In some embodiments, the standalone computing system is at least about 1.5 kW per square foot, at least about 1.6 kW per square foot, at least about 1.65 kW per square foot, at least about 1.8 kW per square foot, at least about 2.0 kW per square foot. , or at least about 3.0 kW of computing power per square foot. In some embodiments, the standalone computing system is about 1.5 kW per square foot or less, about 1.6 kW per square foot or less, about 1.65 kW per square foot or less, about 1.8 kW per square foot or less, about 2.0 kW per square foot or less. , or about 3.0 kW or less of computing power per square foot. It will be appreciated that the height of the standalone system may be adjusted, thereby providing more or less computing power within a given footprint.

It will be understood that the dimensions, components, arrangements, and configurations of the disclosed exemplary embodiments may be modified, added to, and/or subtracted from, to create various potential embodiments in various form factors.

In some embodiments, a standalone computing system, such as a gantry robot 526, configured to remove, replace, and/or install a blade server, power source, or other component, such as the chassis 400, for example. , and may include a robotic system. A standalone system may include an “inside robot” or “outside robot” system. In embodiments with a small footprint, a small magazine 718 of replacement components may be used. In some embodiments, a magazine 718 of replacement components may be attached to the outside of the tank 710 as shown in FIG. 13 . In some embodiments, tank 710 , rack, computer components, power source, replacement magazine, such that gantry robot 526 can remove, replace, and/or install components while moving in substantially only one direction. 718 , and a gantry robot 526 may be arranged. Where several components are arranged substantially linearly, the gantry robot 526 may be moved along a single axis to remove, replace, and/or install the desired component without moving in the second direction. . It will be appreciated that the gantry robot 526 may raise and lower components in addition to moving along a single linear direction.

Using a small form factor, such as the embodiment shown in FIG. 13 , allows a standalone 2PLIC system to be easily transported. The incorporation of a closed loop chiller 719 allows a two-phase liquid immersion cooling system to be used in remote conditions where a substantial source of quenched water may not be accessible. Also, eliminating the need for external cooling water creates, in some embodiments, a standalone computing system that requires only two external connections: a power and data connection.

In some embodiments, the computing system may be housed within an outer housing as shown in FIG. 14 . In some embodiments, the components schematically identified in FIG. 13 and/or disclosed herein may be housed within an outer housing. In some embodiments, the volume of the outer housing may be configured based on expected cooling requirements, the configuration of the closed loop chiller, and/or the environment in which the standalone computing system is expected to be deployed.

The disclosed standalone, self-healing, small form factor embodiments can be used as a standalone solution to provide significant computing power to virtually any location or environment. In some applications, multiple small computing systems may be placed in close proximity to each other, and/or may be linked together to create a cluster. In some embodiments, the outer housing is arranged to allow maintenance and/or service operations to be performed with access to only one side or two sides of the outer housing. This arrangement allows individual standalone computing systems to be deployed with reduced or minimal distances between each standalone system.

In one exemplary embodiment, a cluster of four exemplary standalone computing systems may be strategically positioned to enable about 240 kW of standalone computing power within a footprint of about 140 square feet. In some embodiments, these units may be in power and/or data communication with each other, thereby enabling operation of a multi-unit cluster with only a single external power connection and a single data connection. In some embodiments, a data center may be built using multiple small computing systems or multiple clusters of such computing systems.

While some disclosed embodiments and/or computing systems disclosed herein may be used in modern data centers and/or climate controlled environments, some embodiments of the disclosed standalone computing systems may be deployed in remote locations and/or in harsh environmental conditions. . In some embodiments, the outer housing may be weatherproofed, waterproof, and/or otherwise arranged to withstand exposure to harsh environments for extended periods of time. Some disclosed embodiments allow for rapid deployment of significant computing resources to remote or difficult locations. Some standalone embodiments may be arranged to operate in substantially any position with access to power and data connections. In some embodiments, an uninterruptible power source and/or generator may be operatively coupled to the computing system to provide more reliable and consistent access to power.

Some disclosed standalone embodiments are designed to be stackable. Some stackable embodiments may be designed with reduced height. Certain embodiments may have a height of about 5'5", a depth of 5'6" and a width of 9'. This can result in about 60 kW of computer power in a footprint of 42 square feet. Such units can be stacked vertically, providing 120 kW of computer power within the same 42 square foot footprint.

Embodiments of the disclosed computing system may be stacked, and multiple stacks may be disposed adjacent to each other. This may reduce the need for corridor space between individual computing systems, thereby enabling higher overall power densities within the data center.

Some embodiments may be designed to be fully operational and maintainable, with access to only one aspect of a standalone computing system. Such an embodiment may be advantageous, as it helps to place standalone systems very closely together. Also, in some standalone embodiments, the entire submersion tank may be removed and/or replaced with access to only one side of the device. In certain embodiments, the tanks may be individually modular and/or skid mounted.

In some embodiments, a standalone computing system may be arranged vertically to take advantage of a smaller footprint. A vertical design embodiment of the disclosed system can provide about 60 kW of computing power within an about 22.9 square foot footprint. As in some other disclosed embodiments, some vertically oriented standalone computing systems may be placed close to each other. Also, as described in conjunction with some other embodiments, some vertically oriented standalone computing systems may be operated and maintained with access to only one side of the device. In some embodiments, the entire tank can be removed from the outer housing and replaced. This arrangement allows for rapid replacement of multiple blade servers and/or other computing components.

movable embodiment

A standalone computing system that does not require an external source of cooling water enables new computing applications. In some embodiments, power may be provided to the system using a generator, thereby eliminating the need to connect the system to an external and/or stationary source of power. In some embodiments, the system may rely on wireless data communication.

A fully mobile computing system may be implemented, particularly in standalone embodiments that do not rely on stationary power sources or wired data communications. The disclosed embodiments include a vehicle-mounted, stand-alone computing system that can be utilized to provide significant computing power in virtually any environment. In some embodiments, a truck-mounted, wireless computing system may be carried within the wireless communication range of an existing or temporary network and may provide significant amounts of computing power without significant setup or installation time.

natural water example

In some embodiments, the computing system may be arranged for use on a boat, ship, oil-rig, floating platform, or other vessel or structure located in close proximity to a body of water. In such an embodiment, the condenser used to convert the dielectric vapor back to the dielectric liquid as described herein may be cooled using water from a stream of water. In one exemplary embodiment, the modular computing system may include a water inlet, a water outlet, and a pump or impeller. The pump and/or impeller may cause water to flow from the stream, through the condenser, and then back into the stream. Some embodiments may include filters and/or process components designed to protect condensers, piping, and other computing system components from potential contaminants within a stream of water. In some embodiments, the condenser and other components are arranged to withstand prolonged contact with salt water or brackish water such as sea water, for example.

Horizontal magazine exchange

In some embodiments, a magazine of replacement components may be stored outside of the tank and within an external housing of the computing system. For example, replacement components such as chassis, servers, blades, and/or power components may be removed from the magazine and used to replace components within the tank. The magazine may be positioned on a platform that is configured to extend outside of an outer housing of the computing system to allow components to be replaced from the magazine.

In one non-limiting example, in the event of a malfunction of the blade server within the tank, a robotic arm may be used to extract the non-actuated component from the tank and move the non-actuated component to a storage slot in the magazine. The robotic arm may then remove the functioning blade server from the magazine and install it where the non-operational server was previously installed, thereby replacing the non-operational server with a new operation server.

Over time, the magazine will accumulate non-actuated components that may have been replaced with new working components to allow the robotic system to continue its long-term operation. In some embodiments, the magazine is located on a platform that can extend to the outside of the outer housing, so that an operator can access the magazine. In some embodiments, the platform is configured to rotate the magazine from a substantially vertical position to a substantially horizontal position to enable the component to slide in and out of the magazine.

In some embodiments, an adjustable height cart may be used to move, load, and/or receive components, such that a human operator can lift or receive components while removing or replacing components from the magazine. There is no need to support that weight. It will be appreciated that a magazine configured to rotate to a substantially horizontal position may also assist in loading functional components into the magazine as well as removing non-functional components.

15A-15D show an exemplary magazine 810 positioned on a platform 820 that may extend outside of a container. In FIG. 15A , the magazine 810 may be coupled to a platform including a rotating member 821 , a support member 822 , and a rail 823 . In some embodiments, the support member 822 may be connected to a rail 823 that allows the support member 822 to be moved while supporting the weight of the magazine 810 and any servers or other components stored within the magazine. can In the exemplary embodiment of FIG. 15A , the platform 820 is in an extended position.

15B , during normal operation, the support member 822 may be retracted relative to the outer housing of the computing system. The magazine 810 may be stored on the rail 823 during normal operation. In some embodiments, the weight of the magazine 823 is supported by the support member 822 and the rail 823 , regardless of the position of the support member 822 on the rail 823 .

In some embodiments, computer components used with the disclosed embodiments, such as servers, for example, may be denser and/or heavier than typical computer components. In some embodiments, due to the increased cooling capacity of the disclosed embodiments, the blade server is at least about 50 lbs, or at least about 60 lbs, or at least about 70 lbs, or at least about 80 lbs, or at least about 90 lbs, or at least about It can have a weight of 100 lbs. In some embodiments, the blade server may have a weight of about 50 lbs or less, or about 60 lbs or less, or about 70 lbs or less, or about 80 lbs or less, or about 90 lbs or less, or about 100 lbs or less. As shown in FIG. 15B , the magazine 810 can hold a plurality of chassis 400 or blade servers, and each blade server can weigh about 73 lbs. When the magazine is loaded with three such servers, the combined weight of the magazine 810 and the server may be about 395 lbs.

In some embodiments, the server utilized is a chassis mounted blade server. The server and/or chassis may include a backplane system to facilitate installation and removal of servers within the computing system. In some embodiments, the server may be a submersible server that does not include a fan or other air cooling device. In some embodiments, individual server boards may include 16 GPUs and may be configured to consume about 6 KW of power. In some embodiments, the servers are 1.5U servers. Some disclosed servers may be 1 Otto Immersion Unit (OIU) servers. Such a server is 1.5U high and is configured for liquid immersion cooling. In some embodiments, a single tank in the computing system may be configured to operate at a power of about 60 KW, with 10 10 IU servers and when all 10 servers are operated at substantially full power. In some embodiments, the computing system may include one or two such tanks. In some embodiments, the computing system may include multiple tanks, such as, for example, 10 such tanks.

In some embodiments, as shown in FIG. 15A , when the magazine is extracted from the computing system, the support member is moved from the storage position along the rail and may cantilever outwardly of the outer housing of the computing system.

Also shown in FIGS. 15C and 15D , the magazine may be pulled or otherwise slid outward along the rail and may be cantilevered out of the computing system. In some embodiments, as shown in FIGS. 15C and 15D , the magazine removal tool may be used to remove the entire magazine and components contained within the magazine. In such embodiments, the magazine removal tool may be used to lift the magazine from the support member and slide rails to transport the magazine.

In some embodiments, once the magazine is moved to the exterior of the computing system, the platform may rotate the magazine to a substantially horizontal position. The server housed within the magazine can then be slid out of the magazine.

15A-15D illustrate an exemplary sequence of steps for removing a server from a magazine in accordance with an exemplary embodiment. In an exemplary embodiment, the magazine may be attached to a linear guide rail system behind the access door. 15C and 15D , the magazine can be pulled out of the computing system and cantilevered on the outside. The magazine may be manually pulled out of the computing system or moved out using a motorized or automated system. As shown in FIG. 15D , the magazine may be rotated about 90 degrees to orient the servers and/or other components housed within the magazine in a substantially horizontal position. Once in a substantially horizontal position, the server and/or other component may be slid out of the magazine and onto a cart or other tool configured to receive the server and/or other component. As shown in FIG. 15C , the scissor-elevating cart can be adjusted to a convenient height to accommodate a server or other component. A height adjustable cart with a rolling surface allows the server to be transferred from the magazine onto the cart without the need for a human operator to support the weight of the server. As shown in FIG. 15D , once the server is glided onto a cart with a sliding or rolling surface, the server or other component may be transported anywhere for replacement or service. It will be appreciated that a new component may be loaded into the magazine using substantially the same steps in the reverse order.

In some alternative embodiments, the magazine may be supported on a rotating and extendable arm without rails. In such embodiments, the magazine may be stored in a substantially vertical position within the outer housing of the computing system during normal operation. If it is determined that a component within the magazine needs to be replaced, the magazine may be extended out of the outer housing using an extendable arm. When the magazine extends beyond the outer housing, the magazine can be rotated from a substantially vertical position to a substantially horizontal position, such that components stored within the magazine can be removed horizontally from the magazine.

bellows

In some embodiments, bellows and/or vapor collection systems may be used. Before some disclosed embodiments are initially activated, the dielectric fluid, computer components such as servers, and other system components may be thermally equilibrated. When the computing system is activated, a computer component, such as a server, may begin to generate heat, and such heat may be dissipated into the dielectric fluid. This process causes some dielectric fluids to change from a liquid state to a vapor state. As the temperature of the fluid is increased, a greater proportion of the dielectric fluid can be changed to the vapor phase. In a closed system, an increase in the volume of dielectric vapor can result in increased pressure within the system. In some embodiments, a tank containing the dielectric fluid may be in fluid and/or vapor communication with the recovery system.

16 shows a vapor recovery system 900 according to an exemplary embodiment. The recovery system 900 is connected to a tank 710 containing the dielectric vapor. Dielectric vapor may flow from tank 710 through piping to one or more bellows 905 . In some embodiments, vapor recovery system 900 includes an expanded and collapsing bellows 905 configured to receive dielectric vapor, thereby reducing or eliminating any pressure build-up within tank 710 . As the system cools or when a portion of the dielectric vapor condenses into the dielectric liquid, the bellows may collapse or contract to substantially maintain a pressure equilibrium within the tank 710 .

In some embodiments, the vapor recovery system 900 includes a valve 912 configured to allow ambient air to enter the vapor recovery system. In such an embodiment, the dielectric vapor may be mixed with ambient air. Mixing the dielectric vapor with ambient air can reduce the temperature of the dielectric vapor. In some embodiments, the mixed air/steam may be directed through the carbon bed 911 . The carbon medium in the carbon bed 911 may be configured to draw dielectric vapor while allowing ambient air to pass through the carbon medium and vent from the system 900 through, for example, an outlet valve 913 . can In this way, the heated dielectric vapor can be cooled and captured by the carbon medium.

After operating for a sufficient period of time, embodiments of the computing system will reach a stable thermal state, based on the power capacity of the computing components used. More or less dielectric fluid may be converted to dielectric vapor if more or less computing power is used. This causes the bellows 905 to expand and/or contract in response to heat dissipated into the dielectric fluid.

In some embodiments, bellows 905 may include one or more pouches. Each pouch may include a laminate construction of metal foil and polymer. The bellows pouches may be connected in series or parallel to the vapor recovery system piping and to each other. In some embodiments, the total volume of the inflated bellows pouch may be at least about 15% of the liquid fluid volume of the tank. In some embodiments, the total volume of the inflated bellows pouch may be at least about 20%, or at least about 23%, or at least about 25% or more of the liquid fluid volume of the tank. In some embodiments, the total volume of the inflated bellows pouch may be about 40% or less, or about 30% or less, or at least about 25% or less of the liquid fluid volume of the tank.

In some embodiments, once the computing system has reached thermal stabilization, the vapor recovery system 900 may be closed to cooling ambient air, and a valve allowing air to be vented to the outside of the system may be closed. In some embodiments, the carbon bed may be configured to open only to the tank and bellows using a valve. In some embodiments, a desorption heater configured to transfer heat to the carbon medium may be activated to raise the temperature of the carbon medium. As the carbon medium temperature rises, any dielectric fluid previously captured by the carbon medium may move out of the carbon and back into the tank, where the dielectric fluid will condense back into the dielectric fluid as described above. can

In some embodiments, when power is supplied to the computing system below a previous steady state, a portion of the dielectric fluid in the vapor state may be reduced, and in some embodiments, the bellows may contract to accommodate the reduction in the dielectric vapor. have. In some embodiments, a valve that allows ambient air to enter the bellows is opened, allowing air to enter the bellows and further reducing any pressure differential. In some embodiments, nitrogen may be used instead of ambient air to reduce the pressure differential and also prevent any potential contaminants from being introduced from the ambient air.

In some embodiments, the bellows and/or vapor recovery system may be completely or substantially passive. In some embodiments, the bellows and/or vapor recovery system may be powered and/or automated based on sensor data from temperature, pressure, and/or power sensors deployed throughout the computing system.

In some embodiments, a computing system having a vapor recovery system does not emit emissions, even if the system is not a closed system. In some embodiments, ambient air or nitrogen may be introduced into the system and exhausted out of the system with no or substantially no release of dielectric fluid into the ambient atmosphere.

Exemplary embodiment

The disclosed embodiments enable increased density of computer components and/or computing power. In some embodiments that include a two-phase liquid immersion cooling computer component 170 within a pressure control vessel 110, the components may be separated from each other by less than about 1" or less than about 0.7 inches, or less than about 0.5 inches. In some embodiments, individual components may be separated by greater than about 0.3 inches, or greater than about 0.5 inches, or greater than about 0.7 inches, or greater than about 1 inch, or greater than about 1.5 inches.

Some disclosed embodiments enable improved power utilization efficiency (PUE) compared to conventional data centers. Use of the disclosed embodiment may reduce the use of energy to cool the computer component 170 , thereby reducing the total energy use of the data center and bringing the PUE closer to 1.0. Some embodiments relate to a data center comprising a two-phase liquid immersion cooling computer component within a pressure controlled vessel 110 , wherein the data center is less than about 1.15, or less than about 1.10, or less than about 1.08, or about 1.05. has less than a PUE. Some embodiments relate to a data center comprising a two-phase liquid immersion cooling computer component within a pressure control vessel 110 , wherein the data center is greater than about 1.05, or greater than about 1.06, or greater than about 1.08, or about 1.10. have excess PUE.

In some embodiments, a thermally conductive condensable dielectric fluid is provided for use in a two-phase liquid immersion cooling system. The computer components operate below ambient atmospheric pressure, which lowers the temperature at which the dielectric fluid evaporates, thereby maintaining the liquid phase of the dielectric fluid at a lower temperature relative to standard atmospheric pressure. Computer components generate heat during operation. The heat generated is transferred to the dielectric liquid that comes into contact with the computer components, causing the dielectric liquid to evaporate into a gas. The gaseous dielectric fluid may be condensed using a condenser. Ambient temperature, or quenched process water, is passed through the condenser. When the gaseous dielectric fluid is cooled by the condenser, it condenses back to the liquid phase and falls back into the bath of liquid dielectric fluid.

Some disclosed embodiments relate to high-density data centers. A typical data center contains about 1 megawatt (MW) of computing power distributed over about 10,000 square feet. A high-performance data center may contain about 1 MW of computing power distributed over about 6,000 square feet. The disclosed embodiment relates to a data center comprising a two-phase liquid immersion cooling computer component 170 within a pressure control vessel 110 , wherein the data center is about 3,000 square feet, or about 1,500 square feet, or about 1,000 It utilizes about 1 MW of computing power distributed over square feet, or about 800 square feet, or about 600 square feet. In some embodiments, multiple pressure control vessels containing the disclosed computing system may be arranged in a row and powered by a central power source. In some embodiments, multiple embodiments of the disclosed computing system may be connected in series with each other.

The disclosed embodiment includes a liquid immersion cooling computer component 170 within a pressure control vessel 110 , such that the component is protected from atmospheric contaminants by the pressure control vessel and by immersion in the dielectric liquid 140 . isolate Some disclosed embodiments relate to data centers that operate with minimal air penetration and/or cleaning requirements. In some embodiments, the data center is operated without a HEPA filter or equivalent, or without a MERV 11 filter or equivalent, or without a MERV 8 filter or equivalent.

The disclosed embodiment includes a liquid immersion cooling computer component 170 within a pressure control vessel 110 such that the component is not cooled by gaseous air. Disclosed embodiments include data centers that are operated without cooling fans and/or any other similar devices to circulate air.

Disclosed embodiments relate to environmentally friendly data centers. In some embodiments, the data center includes a liquid immersion cooling computer component 170 within the pressure control vessel 110 and consumes little or no water for the cooling process. Some embodiments use a closed circuit dry cooling tower to cool the condensing structure 130 and reduce the temperature of the water circulated through the disclosed condensing structure 130 to condense the dielectric fluid vapor into a dielectric fluid liquid. Such embodiments operate with no significant input or output of water, since closed loop, dry cooling towers do not rely on streams of water or evaporative cooling for cooling operation. Some data center embodiments utilize less than about 10,000 gallons of water per day, or less than about 1,000 gallons of water per day, or less than about 100 gallons of water per day, or less than about 10 gallons of water per day, or 0 gallons of water per day and / or emit Some data center embodiments use and/or discharge more than about 100 gallons of water per day, or more than about 1,000 gallons of water per day, or more than about 10,000 gallons of water per day.

Disclosed embodiments relate to a computing system comprising: a pressure control vessel operatively coupled to a pressure controller and/or a vacuum source, the pressure control vessel having an interior and an exterior configured to contain an atmosphere within the interior; a volume of a thermally conductive, condensable dielectric fluid; A rack for mounting a computer component, the computer component being arranged to be at least partially submerged in a volume of a thermally conductive dielectric fluid when mounted in the rack; and a condensing structure, wherein the volume of the thermally conductive dielectric fluid, the rack, the computer components, and the condensing structure are contained within the pressure control vessel. Some embodiments relate to a refrigeration system, comprising: a pressure controlled vessel comprising an inner portion, the vessel configured to be operatively connected to a pressure controller for lowering the inner pressure to subatmospheric pressure, the pressure controlled vessel comprising a volume of liquid and a pressure control vessel configured to contain a gaseous thermally conductive condensable dielectric fluid; one or more computer components arranged to be at least partially immersed in a liquid phase of a volume of the thermally conductive condensable dielectric fluid; and a condenser for condensing the gas phase dielectric fluid into the liquid phase dielectric fluid.

In some embodiments, the pressure control vessel is mounted within the upper structure, the blade server is configured to be interchangeable without interfering with the computing system, the pressure control vessel is operatively connected to a power source, a source of water, and networking connections, the pressure control vessel is The vessel includes an opening in the upper end and a lid configured to sealably close the opening, the lid configured to direct rising vapor from the middle of the pressure control vessel to a side of the pressure control vessel, the pressure control vessel having about 100 cubic feet to about 300 cubic feet, and/or the pressure control vessel comprises a ratio of liquid dielectric fluid to gaseous dielectric fluid of from about 1:3 to about 1:8. Some embodiments further include a ballast block, a blade server and blade server chassis, a robot arm and an airlock, and/or a purge system, wherein the airlock does not substantially disturb the atmosphere within the pressure control vessel, the pressure control vessel and the purge system is configured to remove contaminants from the volume of the thermally conductive dielectric fluid. In some embodiments, the purge system is configured to remove a portion of the atmosphere from the pressure control vessel, condense any dielectric fluid from the atmosphere, and discard any remaining vapor. In some embodiments, the purge system is configured to condense at least a portion of the gaseous dielectric fluid and dispose of the gaseous contaminants.

Some embodiments relate to a method for cooling a computer component, the method comprising: providing a housing, the housing comprising a thermally conductive condensable dielectric fluid and a heat-generating computer component, the housing comprising at least slightly configured to withstand a vacuum of; operating a computer component, wherein the operation of the computer component generates heat and the computer component is in contact with a dielectric fluid; and creating a vacuum within the housing, wherein the pressure within the housing is at least less than about 1 atmosphere. Some embodiments include maintaining a vacuum within the housing, wherein the pressure within the housing is less than about 1 atmosphere during operation of the computer component; using heat generated by the computer component to transfer the dielectric fluid from a liquid state to a gas. and condensing the dielectric fluid from the gaseous state to the liquid state using a condenser, and removing the non-condensable fluid from the dielectric fluid. In some embodiments, and/or replacing some of the computer components while the system is in operation. In certain embodiments, removing the non-condensable fluid comprises isolating a portion of the gaseous atmosphere from within the housing, condensing any dielectric fluid from the gaseous atmosphere; returning the condensed dielectric fluid to the housing, and discarding any remaining portion of the gas atmosphere, and/or the housing is configured to create convection.

Some embodiments relate to a method of cooling a computer component comprising operating the computer component below ambient pressure, wherein the computer component is contacted with a thermally conductive dielectric fluid. Some embodiments further include evaporating the dielectric fluid and condensing the dielectric fluid below ambient pressure.

Some embodiments relate to a method for cooling a computer component, the method comprising: providing a thermally conductive condensable dielectric fluid in a liquid phase and a gas phase; and operating the computer component at a pressure below ambient atmospheric pressure in the presence of the thermally conductive condensable dielectric fluid, wherein the computer component is at least partially in contact with the thermally conductive condensable dielectric fluid in the liquid phase. Some embodiments include evaporating a dielectric fluid from a liquid phase to a gas phase using at least a portion of any heat generated by the computer component being operated on; condensing at least a portion of the dielectric fluid from the gas phase to the liquid phase; removing from the dielectric fluid at least a portion of the fluid that cannot readily condense; and/or replacing at least one or more computer components while the computer components are being operated.

Some embodiments relate to a method of cooling a computer component, the method comprising: operating the computer component at at least 1 psi below ambient pressure, the computer component comprising: a thermally conductive dielectric fluid and at least Partially contacted, the boiling point of the dielectric fluid is less than about 80°C. Some embodiments further include condensing the dielectric fluid under conditions such that the computer components do not exceed about 80°C.

It will be understood that the various disclosed embodiments may include some or all of the elements otherwise described herein. Certain components and their characteristics may be adjusted based on the characteristics of each particular embodiment. Modifications may include the use of higher or lower density power, cooling and network connectivity systems, pressure management systems, steam management systems, and selection of special equipment and components.

From the foregoing, those skilled in the art can readily ascertain the essential nature of this disclosure without departing from the scope and spirit of the disclosure: various changes and modifications may be made to adapt the disclosure to various uses and conditions. There will be. The foregoing embodiments are meant to be illustrative only, and should not be construed as limiting the scope of the disclosure.

Heating and cooling of tanks in response to shock events

In one exemplary embodiment, an immersion cooling system or vessel may include a tank, a computing device, a robot, an absorption unit, a bellows and a management system. The tank may be a pressure controlled tank maintained at (or within) atmospheric pressure. The tank may include a bath area and a sump area, and the computing device may be immersed in a dielectric fluid within the bath area of the tank. The computing device may be connected to a network and may perform various processing tasks while immersed in the dielectric fluid. The tank may include a cover for accessing the bath area, computing device, and sump area. The tank may be fluidly coupled to the bellows and absorption unit, and a plurality of valves may selectively connect or disconnect the tank to or from the bellows and absorption unit, such that the dielectric vapor is delivered to the bellows and/or absorption unit. can be, or vice versa. The robot may be a gantry robot capable of lifting the computing device from the tank of the vessel when the lid of the tank is opened. The robot may place the elevated computing device within a magazine provided for storage of the computing device. The robot may also raise the computing device from the magazine and place it in place of the elevated computing device from the tank.

In one exemplary embodiment, the tank may include a heating element, eg, a plurality of heating rods, a portion of the heating rods being at least partially immersed in the dielectric fluid. The tank may include a plurality of sensors, such as a temperature sensor, a pressure sensor, or sensors, that provide operational data (eg, current, voltage, workload, etc.) related to the computing device. The temperature sensor may be located inside the bath or in an area above the bath. Using the data received from the sensor, the management system of the vessel may operate the heating element to regulate or control the temperature or temperature fluctuations (and/or pressure or pressure fluctuations of the dielectric vapor) of the dielectric fluid within the tank. 18 shows an example of a heating element 1000 for an immersion cooling system according to an exemplary embodiment. The heating element 1000 may include a plurality of heating rods 1010 . Each heating rod may include a plurality of wires 1011 that may be connected to a power source in the tank. A controller of the tank may adjust the heating element 1000 to heat the bath region of the tank, for example during various operations of the tank. In this exemplary embodiment, the heating element 1000 may be mounted within the bath of the tank and may be completely submerged in the dielectric fluid.

In one exemplary embodiment, the heating element is separate from the computing device, and the heating element does not process data. A heating element can be specified to only generate heat and not to function otherwise. The heating element can be easily controlled, particularly during operation of the tank (eg, starting operation), replacement of components, or other times when control is required. The heat generated by the heating element can be adjusted when evaluating data indicative of the size of the bellows and other aspects of the system, such as pressure or temperature.

In particular, an abrupt change in the power consumption or workload of a computing device (eg, caused by or in the absence of end-user activity) may result in an abrupt change in the amount of heat generated by the computing device within the vessel. have. This in turn can lead to rapid temperature changes in the tank or bath, which can lead to rapid changes in the pressure in the tank (in a closed insulation system, since pressure and temperature are directly related, i.e. PV = nRT) . Such pressure fluctuations can damage the vessel and introduce contaminant gases (eg air) or particulates (eg dust) into the tank. These pressure fluctuations can also cause leakage of dielectric fluid from the tank. To counteract the effects of these pressure fluctuations, a bellows or absorption unit may be used to remove excess vapor from the tank or introduce vapor into the tank as the pressure drops. However, by using a heating element, the capacity of the bellows and absorption unit can be reduced, whereby a more space efficient container can be designed. If a heating element is not used, in the event of an excessive increase in pressure, the bellows may rupture.

The heating element makes it possible to regulate changes in temperature within the tank or bath, thereby facilitating a controlled transition between various operating load conditions that may occur in the computing device during operation. For example, in the event of a sharp decrease in the operating workload of the computing device, the heat generated by the computing device may drop sharply. This can cause a sharp decrease in the internal pressure of the tank. The heating element may add heat to the tank such that a decrease in the temperature of the dielectric fluid is controlled, eg, during a shutdown process. In other words, the heating element may balance the pressure and temperature of the tank in the event of a sudden change in the workload of the computing device, ie, in the event of a shock. Thus, the vessel requires a much smaller bellows and absorption unit to maintain the atmospheric pressure of the tank.

In one exemplary embodiment, the management system of the vessel may determine how much heat to add to the tank in response to a shock event, eg, an increase or decrease in internal pressure or temperature of the tank. In one exemplary embodiment, the rate of drop (or increase) in temperature or pressure may determine how much heat to add to the tank. For example, in the event that the dielectric fluid temperature level of the tank decreases by more than a certain number of degrees, over a certain number of minutes, the management system can activate the heating element to add a certain amount of heat to the tank. (e.g., to maintain the temperature and pressure of the system). This added heat can stop the decrease in temperature, or it can slow the rate at which the temperature falls. The management system may stop the heating element from adding heat to the system when the tank is in a steady state, eg, when the rate at which pressure or temperature drops is below a threshold. In another exemplary embodiment, the actual temperature of the dielectric fluid in the tank may determine how much heat it will add to the tank when an increase or decrease in the workload of the computing device is initiated.

In one exemplary embodiment, the management system may activate the heating element when a shock event is detected, eg, before, during or after a starting operation, boosting operation, decelerating operation, or stopping operation. . The management system may, by receiving sensor data (eg, from a temperature or pressure sensor within the tank) or data from a computing device (eg, current, voltage, temperature, workload, data transfer, etc.), of the operating mode (eg, starting or stopping). The heating element may adjust or regulate changes in temperature or pressure within the tank to minimize deviations in pressure from atmospheric pressure. Alternatively, in the absence of operation of the heating element in accordance with the techniques disclosed herein, the container must absorb or store excess gas produced as a result of the rapid heating of the computing device, or the container may cause a sharp drop in heat production of the computing device. Gas will have to be desorbed or supplied to counteract the pressure drop as a result of

During a startup operation, the temperature of the tank, eg, the temperature of the dielectric fluid in the bath, is below a threshold when the computing device starts operation. A start-up operation may occur, for example, when the vessel has just been turned on and the tank is cold. Because the computing device heats up rapidly, the computing device can generate a lot of steam when the dielectric fluid is cold. Accordingly, prior to, during or after the startup operation, the management system may activate the heating element to heat the dielectric fluid, thereby controlling the temperature of the dielectric fluid and minimizing vapor production by the computing device. can For example, the heating element may slowly increase the temperature of the dielectric fluid to a threshold temperature prior to turning on the computing device. Otherwise, in order to maintain the tank at atmospheric pressure, the vessel will have to contain an excess of vapor, which may require large capacities in the bellows and absorption units.

During a boost operation (eg, when the temperature of the tank is below a threshold), the temperature of the tank, eg, the temperature of the dielectric fluid in the bath, may increase faster than the threshold rate. A boost operation may occur, for example, when the computing device is being operated and the workload of the computing device is significantly increased, for example, due to an increase in customer demand. A sharp increase in the workload of the computing device may increase the amount of heat generated by the computing device, thereby increasing the amount of steam generated by the computing device. Accordingly, prior to, during or after the boost operation, the management system may activate the heating element to heat the dielectric fluid, thereby controlling the temperature of the dielectric fluid and minimizing vapor production by the computing device. can Otherwise, in order to maintain the tank at atmospheric pressure, the vessel will have to contain an excess of vapor, which may require large capacity for storage and absorption in the bellows and absorption unit.

During a shutdown operation (eg, when the temperature of the tank is above a threshold), the temperature of the tank, eg, the temperature of the dielectric fluid in the bath, may decrease faster than the threshold rate. An interruption operation may occur, for example, when the computing device is operating and the workload of the computing device is significantly reduced, for example, due to a decrease in customer demand. A sharp reduction in the workload of the computing device may reduce the amount of heat generated by the computing device, thereby drastically reducing the pressure in the tank. Accordingly, before, during or after the abort operation, the management system may activate the heating element to heat the dielectric fluid, thereby reducing the temperature of the dielectric fluid in a controlled manner and minimizing the pressure drop in the tank. . Otherwise, in order to maintain the tank at atmospheric pressure, the vessel would have to produce an excess of vapor, which may require large storage or desorption capacity in the bellows and absorption units.

During a shutdown operation (or controlled shutdown process), while the temperature of the tank, eg, the temperature of the dielectric fluid in the bath, is above a threshold, the vessel is ordered to turn off. Because the computing device can abruptly stop generating heat, the pressure in the tank can drop sharply. Accordingly, prior to, during or after the interruption operation, the management system may activate the heating element to heat the dielectric fluid, thereby reducing the temperature of the dielectric fluid in a controlled manner and minimizing the pressure drop. For example, the heating element may slowly heat the dielectric fluid such that the temperature of the dielectric fluid slowly drops when the computing device is turned off. Otherwise, in order to maintain the tank at atmospheric pressure, the vessel would have to produce an excess of vapor, which may require large storage or desorption capacity in the bellows and absorption units.

In one exemplary embodiment, as the vessel responds to a shock event, the management system (or other system) may add vapor or fluid to the tank or vapor to maintain the tank's pressure at near-atmospheric pressure. Alternatively, the fluid may be removed from the tank. For example, vapor or fluid may be removed from the tank as the temperature of the tank is increased, and vapor or fluid may be added to the tank as the temperature of the tank is decreased.

The vessel may utilize a variety of mechanisms to add vapor or fluid to the tank or to remove vapor or fluid from the tank. In one exemplary embodiment, the vessel may utilize a bellows as a mechanism for adding vapor to or removing vapor from the tank. In another exemplary embodiment, the vessel may utilize an absorption/desorption unit (hereinafter, “absorption unit”) to add vapor to the tank or to remove vapor from the tank. In another exemplary embodiment, the vessel may utilize a pressurized container to add vapor to or remove vapor from the tank. In another exemplary embodiment, the vessel may utilize a combination of the mechanisms listed above to add vapor or fluid to the tank or to remove vapor or fluid from the tank. In another exemplary embodiment, the vessel may maintain the pressure of the tank using a combination of a heating element and one or more of the mechanisms listed above.

For example, during a start-up operation, the management system may use a combination of a heating element and a bellows to maintain the pressure in the tank. In one example, prior to turning on the computing device, the management system may activate the heating element to heat the dielectric fluid. At some point (eg, before, after, or during heating), the management system may open a valve connecting the bellows to the tank, thereby facilitating delivery of dielectric vapor to the bellows. This delivery of dielectric vapor to the bellows may prevent an uncontrolled increase in tank pressure, thereby increasing the temperature of the dielectric fluid while the pressure in the tank may be maintained (eg, within a tolerance range). make it possible

Similarly, during a start-up operation, the management system can maintain the pressure in the tank using a combination of a heating element and an absorption unit. At some point (eg, before, after, or during heating), the management system may open a valve connecting the absorption unit to the tank, thereby allowing the dielectric vapor to absorb or retain the dielectric vapor within the absorption unit. It may facilitate transfer to an absorbent unit, for example, a carbon bed. Similarly, during a start-up operation, the management system may maintain the pressure in the tank using a combination of a heating element and a pressurized container. At some point (eg, before, after, or during heating), the management system may open a valve connecting the pump and pressurized container to the tank, thereby using the pump to deliver the dielectric vapor to the pressurized container. can promote The pressurized container may store dielectric vapor.

As another example, during a shutdown operation, the management system may use a combination of a heating element and a bellows to maintain the pressure in the tank. In one example, after turning off the computing device, the management system may activate the heating element to heat the dielectric fluid. At some point (eg, before, after, or during heating), the management system may open a valve connecting the bellows to the tank, thereby facilitating delivery of dielectric vapor to the tank. This delivery of the dielectric vapor to the tank may prevent an uncontrolled decrease in tank pressure, whereby the temperature of the dielectric fluid may be reduced while the pressure in the tank can be maintained (eg, within a tolerance range). make it possible

Similarly, during shutdown operations, the management system may utilize a combination of heating elements and absorption units to maintain pressure in the tank. At some point (eg, before, after, or during heating), the management system may open a valve connecting the absorption unit to the tank, thereby facilitating delivery of the dielectric vapor to the tank. In the case of a carbon bed as an absorption unit, the management system can activate the carbon bed to release trapped or absorbed dielectric molecules. The management system may activate the carbon bed, for example, by sending a signal to a switch to turn on a heating device in the carbon bed. In one example, as the pressure in the tank drops, the carbon bed can be heated to release the dielectric vapor and minimize the pressure drop.

Similarly, during a shutdown operation, the management system may use a combination of a heating element and a pressurized container to maintain the pressure in the tank. At some point (eg, before, after, or during heating), the management system may open a valve connecting the pressurized container to the tank, thereby facilitating delivery of the dielectric vapor to the tank.

In one exemplary embodiment, a compromise may be made between the use of bellows and the use of an absorption unit. The bellows is a passive device, while the absorption unit is an active device. Bellows-based systems can be more power efficient than absorption-unit-based systems, since bellows do not require active heating. However, bellows occupy more space than absorption units, and absorption-unit-based systems provide a greater degree of control and functionality. In this regard, efficiency, control and space may be some of the design constraints.

In one exemplary embodiment, an uncontrolled interruption in the vessel may occur. For example, due to a loss of power, an uncontrolled shutdown in the vessel may occur. In this exemplary embodiment, an emergency shutdown process may be implemented to address possible pressure fluctuations within the tank. For example, a vessel may have a backup or uninterruptible power supply (“UPS”) capable of powering the vessel and its management system (or other system). When the management system receives a signal from a sensor indicating that the pressure in the tank is dropping below an acceptable threshold as a result of the loss of power and cooling of the system, it may instruct the management system's bypass valve to open. A bypass valve may connect the tank to the environment outside the tank. The bypass valve may introduce air into the tank and thus normalize the pressure in the tank (so the tank or bellows will not collapse). Thereafter, during the start-up operation, the vessel may purge the air introduced into the tank.

In one exemplary embodiment, the management system (or other system) may use a table, matrix or map (“map”) to determine how to respond to a shock event. In one exemplary embodiment, the map may display a change in temperature as an input and may display as an output how much heat to add to the tank in response to the change in temperature. In one exemplary embodiment, the map provides as inputs data related to vapor temperature, tank pressure, fluid level in a tank or sump area, fluid pressure in a pump or filter, pressure differential, humidity level, and alumina condition. may include In response to these inputs, the map may provide outputs such as condensers, heating elements, pumps, bellows valves, carbon injection valves, carbon exhaust valves, and computing device operating parameters. The map may define various states for the operation of the vessel. The management system may receive various data from sensors provided throughout the container. Using the map, the management system can transform such data into operational parameters for the devices in the vessel, such as bellows, absorption units, valves, heating elements, pumps, condensers, and computing devices.

In one exemplary embodiment, the vessel may be operated at a temperature near the boiling point of the dielectric fluid and a pressure near atmospheric pressure. However, one of ordinary skill in the art will recognize that the vessel may be operated at other temperatures and pressure ranges, based on the optimum operating temperature for operation of the computing device. In one exemplary embodiment, the optimal operating temperature of the system is about 137°F ± 8°. In one exemplary embodiment, the optimal operating pressure of the system is approximately atmospheric pressure (eg, 101,325 Pa) ± 5,000 Pa. In this exemplary embodiment, during a shock event, the management system will attempt to maintain the temperature and pressure of the vessel within these ranges.

Although, in some exemplary embodiments of this disclosure, the management system is described as a system programmed to perform various tasks in a shock event, those skilled in the art will recognize that other systems disclosed in this disclosure may be programmed to perform such tasks. will recognize that

In one exemplary embodiment, the vessel can be operated under three modes of operation. In a first mode of operation, the tank may be operated at atmospheric pressure. In a second mode of operation, the tank can be operated at a pressure range significantly deviating from atmospheric pressure. In a third mode of operation, the vessel may sometimes be operated at atmospheric pressure, and sometimes at a pressure range significantly deviating from atmospheric pressure. The third operation mode may be a combination of the first mode and the second mode. In one exemplary embodiment, the management system may determine the mode of operation of the container. For example, the management system may operate the container based on rules prescribed for the management system, for example, pressurize the container at 5 AM every morning and return to atmospheric pressure at night; The vessel can be pressurized during peaks of the workload determined by the sensor data. As another example, the management system may use a machine learning algorithm to predict the mode of operation for a vessel. For example, the machine learning algorithm may use not only sensor data but also external data, for example, climatic conditions, calendar data, usage data, and the like, to predict which operation mode is most efficient under a corresponding situation. A user of the system can provide the management system with labeled data from which the data can be extrapolated to create a model for prediction of the mode of operation.

In one exemplary embodiment, the management system may carry out certain routines before the lid of the tank may be opened. For example, if the vessel is instructed to open the lid of the tank, the condensing system may allow the system to cool for a period of time before the management system allows the lid to be opened. The condensing system may minimize vapors in the tank and thus may minimize loss of dielectric vapors to the environment when the lid is opened.

In one exemplary embodiment, the immersion cooling system may be a modular system. For example, each group of components of the system may be mounted on a separate skid, such as a condensing skid, a heating skid, a bellows skid, an absorption unit skid, and the like. These skids can be moved and deployed for a variety of applications.

Dielectric Fluid Circulation and Filtration

In one exemplary embodiment, the vessel may include a pump for circulating the dielectric fluid through the tank. For example, a tank may include a sump area and a bath area. The bath region may retain the computing device immersed in the dielectric fluid. The sump region may be located next to the bath region, or the sump region may be in fluid communication with the bath region. For example, the sump region may receive an overflow of dielectric fluid from the bath region, eg, the dielectric fluid may flow over the walls of the bath region adjacent the sump region. A pump may draw dielectric fluid from the sump region and pass such fluid through a filter. After the filter, the dielectric fluid may be returned to the bath area. The vessel may include various pipes coupling the sump region, pumps, filters, and bath regions.

In one exemplary embodiment, an amount of dielectric fluid is provided to the vessel such that the bath region is filled with the dielectric fluid and there is an overflow of the dielectric fluid within the sump region. The full bath area ensures that the computing device is completely immersed in the dielectric fluid. The pump may draw the dielectric fluid from the sump region and deliver it, for example, through a filter, to the bath region. Because there is more dielectric fluid in the tank than the bath area can hold the fluid, the bath area is always full when the pump is running (especially when the pump is running). However, depending on the temperature of the tank, the level of dielectric fluid in the sump region may change because the dielectric fluid may evaporate from the bath region and the dielectric fluid from the sump region may displace the evaporated fluid in the tank. am.

In one exemplary embodiment, the tank may be in the shape of a rectangular box. The dielectric fluid may flow beyond the upper end of one of the short sides into the sump region adjacent the short side. The pump can draw in a dielectric fluid and use such dielectric fluid to minimize disruption or induce turbulence of the fluid in the bath (because turbulence or turbulence can cause cavitation in the fluid). ), can be returned to the position in the tank or reintroduced. In particular, the greater the distance between the overflow area and the re-introduction point, the less turbulence associated with the re-introduction of the fluid into the tank. For example, if the dielectric fluid overflows from the upper end of the first side of the bath, the pump may return the dielectric fluid to the lower end of the side opposite the first side. The pump may return the dielectric fluid to the edge of the bottom side, which minimizes mixing or turbulence of the fluid in the bath.

In one exemplary embodiment, the vessel may include two pumps. Each pump can independently draw fluid from the sump area and deliver it to the bath area. Providing a vessel with two separate and independent pumps can improve the service life of the vessel. Also, if one of the pumps fails for any reason, the vessel can continue its operation without interruption until the failed pump is replaced.

In one exemplary embodiment, the container may include a filter. A filter may include one or more cores. Each core may filter the dielectric fluid for different type(s) of contaminants, particles, substances, diluents or solutes. In one exemplary embodiment, the core may be selected based on the characteristics of the dielectric fluid and contaminants likely to be introduced into the dielectric fluid. For example, contaminants may include solders and resins used during the manufacturing process of electronic boards used in computing devices. The dielectric fluid can act as a cleaner for resin, solder, dust, dirt or anything else in the system. Solder and resin (or other material) can be washed out of the electronic board after it is immersed in the dielectric fluid. The filter can remove solder and resin (or other material) from the dielectric fluid. If such material is not removed from the dielectric fluid, as the dielectric fluid evaporates, such material will deposit as a layer on the heat generating component of the computing device, eg, a processor. Consequently, such a layer thermally isolates or insulates the heat generating component from the dielectric fluid, thereby reducing the efficiency of heat transfer from the component to the dielectric fluid. Accordingly, the component may be heated and destroyed more frequently.

In one exemplary embodiment, the filter may include two cores, one core may include activated carbon (charcoal) and the other core may include activated aluminum. For example, the ratio of activated carbon to activated aluminum may be 3 to 1. As another example, the filter may include four cores, three cores may include activated carbon and one core may include activated aluminum.

In one exemplary embodiment, the filter may include a stripe for testing the acidity of the dielectric fluid. The stripe may be a PH indicator, litmus paper, or other indicator. In some cases, the dielectric fluid may become acidic after interacting with certain components of the tank. The stripes may be contacted with the dielectric fluid, and if the dielectric fluid is acidic, the color may change. The filter may include a color detection sensor, the color detection sensor may detect a change in color in the stripe, and when a color change in the stripe is detected, may send a signal to a management system (or other system). In one exemplary embodiment, the stripes may be disposed within a container or chamber comprising a glass shield. Therefore, the color change of the stripe can be seen from the outside of the container. A camera may be placed within the proximity of the container. The camera can take pictures of the stripes (behind the glass shield) and send those pictures to a management system. If the management system (or the user of the system) detects a color change in the stripe (using data provided by the camera or color sensor), the management system can trigger a remedial action, for example a maintenance system may be notified or the system may be shut down.

In one exemplary embodiment, the camera may be a pan-tilt-zoom camera. A filter cover may be mounted on the top of the sump area. A filter shroud may be installed next to another shroud that provides access to the bath area. The filter cover may include a filter, and the camera may be installed on the filter cover. In one embodiment, the camera may be installed directly under the filter cover. Thus, when the camera is rotated, the camera can capture the shape of the stripes, the sump area (the area under the camera) and the area above the bath area.

19A and 19B show a filter comprising three cores according to an exemplary embodiment. As shown in FIG. 19A , the filter may include a lid 1050 that may be mounted to the tank, for example, followed by another lid that provides access to computing devices installed within the tank. Each core of the filter may be connected to the lid 1050 . Lid 1050 may include three caps 1060, each cap providing access to one of the cores. 19B shows structure 1070 mounted to lid 1050 . Structure 1070 may support various filter cores and other components, such as filter core 1071 , camera 1072 , and electromechanical valve 1073 . On the other side of the structure 1070, there may be two other filter cores (not shown in FIG. 19B ).

In this exemplary filter, a camera and two color sensors are attached to the cover. The camera and color sensor may acquire data related to the acidity of the dielectric fluid (based on the color of the stripes) and communicate such data to a management system.

In one exemplary embodiment, the filter may be mounted on a chassis that may be removed by a robot. The chassis may include a connection interface for removably connecting the chassis (and filters provided therein) to various pipes provided within the tank. Thus, when the management system determines that the filter needs to be replaced, the robot can raise the chassis out of the tank and place the filter within the magazine.

In one exemplary embodiment, the management system may notify the user when service or replacement of the filter is required. For example, the management system may include a timer or counter that is activated when the filter is installed in the container. If the management system determines that the filter has been operated for longer than a threshold time, the management system may send a notification to the user (or other entity). As another example, the management system can activate a timer or counter only when the container is running, when the pump is active, or when dielectric fluid passes through the filter (as determined by a fluid sensor in the filter). have. If the management system determines that the filter has been operated for longer than a threshold time, the management system may send a notification to the user. As another example, the management system may determine a pressure differential for the filter, and the management system may notify the user of service or replacement of the filter when the pressure differential exceeds a threshold pressure. In particular, the filter may include an input pipe and an output pipe, and there may be a pressure sensor on the input pipe and a pressure sensor on the output pipe. Each pressure sensor may send a pressure reading to the management system. If the pressure difference between the readings of the pressure sensors exceeds a threshold pressure, the management sensor can determine that the filter is clogged. Accordingly, the management system may notify the user of service or replacement of the filter. As another example, the filter may include a sensor indicating a flow rate to the filter. The management system may use the flow rate to determine if the filter needs servicing. As another example, the management system may use a machine learning model to determine when to replace a filter. Such a model may receive training data from a central server representing operational data for filters for a plurality of vessels coupled to the server.

In one exemplary embodiment, the sump area and/or the bath area may include one or more fluid level sensors. During startup or rapid increase in workload, the fluid level in the sump region is lowered as the dielectric fluid evaporates within the bath region. However, as the pump circulates the dielectric fluid, the fluid level in the bath region remains the same, ie the computing device remains submerged. Conversely, during interruptions or sharp reductions in workload, the fluid level in the sump area may decrease.

The fluid level sensor may provide data related to the fluid level in the sump area and the bath area to the management system. If the fluid level in the sump area decreases after starting the vessel (or during normal operation of the vessel), there may be a leak in the tank. Similarly, if the fluid level in the bath region decreases at a certain point, there may be a problem with the fluid circulation system, for example a pump failure. Accordingly, the management system may continuously monitor the fluid level data provided by the fluid level sensor and notify the user in the event of an unexpected drop in fluid level within the sump area or bath area.

In one exemplary embodiment, a pump may draw fluid from a sump region (or bath region) and may provide fluid to a drain valve connected to the tank body. When the valve is opened, the pump may drain the sump area (or bath area) or provide a sample to the user of the vessel. A sample may be provided to a laboratory for testing. In one exemplary embodiment, the user, using the management system, may instruct the vessel to drain the tank. In response, the management system may open the drain valve and the pump may direct fluid from the sump area (or, for example, when a connection is provided to the bath area, even the bath area) to the drain valve. For example, there may be a valve connection between the bath region and the sump region, and in an event that a drain instruction is received, the valve connection may connect the bath region to the sump region, such that the bath region directs the dielectric fluid into the sump region. Drain, and the pump drains the sump area. In one exemplary embodiment, the pump may draw the dielectric fluid directly from the tank area.

robot system

In one exemplary embodiment, the vessel may include a robotic system, eg, a gantry robot configured to raise a computing device from a magazine disposed proximate to or in the bath area of the tank. The gantry robot may also place the computing device within the bath area or magazine.

A gantry robot (or robot) may include a series of linear actuators. For example, the robot may include actuators for movement along each of a plurality of directions, eg, horizontal and vertical directions. A management system (or other system) may control how far or how fast each one of these actuators will move. In one exemplary embodiment, the actuator may be configured to move over one or more tracks. Actuator-based or track-based systems may lose their accuracy over time (eg, due to drifting or wear and tear). Thus, in this exemplary embodiment, the tank (or various components thereof) may include one or more calibration zones or flags in order to detect the exact relative position of the robot. For example, one or more of the critical components or locations of the container with which the robot interacts, such as a magazine, a first server rack, a second server rack, or a home position, may include a flag; Such flags may be detected by the robot after the robot has reached the location or location of the critical component. The flag can notify the robot of the robot's exact position relative to a critical component or place.

In one exemplary embodiment, the flag may be a physical object, an RFID tag, a color, an alphanumeric code, a QR code, or the like. In one exemplary embodiment, the sensor detecting the flag may be a motion sensor, an RFID detector, a camera, or the like. In one exemplary embodiment, the camera may determine the distance between the robot and various objects, and may provide feedback related to the distance to the robot. In one exemplary embodiment, the camera may provide video data to the management system, and based on the video data, the management system may determine the exact location of the robot within the container. In one exemplary embodiment, the management system may determine the position of the robot, for example by scanning a QR code, counting the components in the tank, or the like. In one exemplary embodiment, images from the camera may be used to determine the proximity of the robot to an object or whether the robot has properly gripped or positioned the chassis. In one exemplary embodiment, the management system may use object recognition technology to determine the position of the robot. In one exemplary embodiment, the management system may use artificial intelligence technology to determine the position of the robot. The management system can use object recognition technology or artificial intelligence to calibrate the robot.

In one exemplary embodiment, the container may include a home location, a magazine and two racks. The management system (or other system) may instruct the robot to lift the computing device, for example, from the second rack. The robot can move from the home position to the magazine and then to the first and second racks. As the robot approaches each of these locations or components, the robot's sensors may detect an associated flag for the location or component. An advantage of the flag system is that the robot can still detect the position of the robot with respect to the critical component or position even when other components or positions have been removed from the container. This is because flags are always located in the same position for each critical component or position associated with that flag. For example, even when the first rack is removed from the container, the robot can find the flag for the second rack, calibrate the robot's position relative to the second rack, and remove the computing device from the second rack. can do. Similarly, even if the second rack has been moved slightly from its position within the container, the robot can find the flag for the second rack, calibrate the robot's position with respect to the second rack, and move the computing device to the second rack. It can be removed from the rack.

In one exemplary embodiment, the gantry robot may receive instructions to remove or replace various components of the container, such as computing devices, filters, and the like. In one exemplary embodiment, such instructions may be provided by a management system (or other system). The management system may provide instructions in response to a determination by the management system (or other system), a user of the container, or a system external to the container. For example, the management system may receive and monitor various data points related to the operation of the computing device, such as voltage levels, temperatures, and other operating characteristics. If the computing device exceeds a determined or predetermined threshold for the computing device, the management system is programmed to instruct the robot to replace the computing device.

As another example, a user of the container may instruct the management system to provide instructions to the robot to remove the computing device. As another example, the management system includes an application programming interface (API) for receiving instructions from a system external to the container. For example, the vessel may communicate with a top-level organization and management platform that may instruct the management system (via API) to remove the computing device from the tank.

In one exemplary embodiment, the robot may raise the computing device from a tank or magazine. In this exemplary embodiment, the computing device may be located within a chassis that includes a connecting plate. The robot may include a linkage mechanism and guide pins that may interface with the connecting plate. The robot may also include one or more load cells that measure the positive or negative force or pressure applied to the robot.

The robot can start from its home position and can be moved towards the tank (or rack containing the computing device). At the tank, the robot may detect a flag associated with the tank, which may inform the robot that the robot is located in the tank. The robot may then move a predetermined distance from the flag so that the robot is accurately (or substantially) positioned over the computing device. When the robot is positioned on the top of the computing device, the robot moves from its top position to a location a few inches away from the computing device (or a location equal to or greater than the length of the guide pin, for example 50% further than the guide pin). can descend quickly. At such a point (ie, a few inches away from the computing device), the robot may approach the computing device more slowly, such that the robot's guide pins are initially in contact with the connecting plate. Once the robot makes initial contact, it will continue to move downward at the same slow speed until the robot presses the connecting plate (of the chassis) above the threshold pressure. At that point, the robot's linkage mechanism may be activated (eg, fingers may be opened), thereby interconnecting the robot to the computing device. The engagement mechanism may be an armature-based engagement mechanism comprising a plurality of fingers. When the linkage mechanism is closed, the robot can provide feedback to the management system that the linkage system is closed. The management system may instruct the robot to raise the computing device. The robot can slowly pull the chassis upward a few inches, thus confirming that it has a good grip on the chassis. The robot can then quickly move upwards to its topmost position. From that point, the robot can move to any location dictated by the management system, for example a magazine or other rack.

In one exemplary embodiment, the robot may place the computing device within a tank or magazine. For example, while holding the chassis, the robot can move over a slot in the tank (or one of its racks) and drop or place the chassis into the tank (or magazine). Once the robot is positioned over the tank, it begins to move rapidly downward until it is positioned a few inches above the first alignment point (or mating point) between the chassis and the tank (or rack). The design of the chassis and tank can determine the distance from above the tank the robot must slowly descend. In particular, the robot can be lowered slowly to an inch or two above the alignment point (where the guide rails of the chassis contact the grooves in the rack). The robot can move slowly towards the tank so that the grooves in the rack can move within the guide rails of the chassis. The management system can receive and monitor data from load cells and other sensors, thereby ensuring that the chassis is not misaligned. For example, feedback of an oversized force on the load cell may indicate a misalignment between the groove and the guide rail. If misalignment is detected, the management system may abort the lowering operation.

In one exemplary embodiment, in addition to the grooves and guide rails, the chassis and rack may include additional alignment mechanisms. For example, after initial occlusion between the chassis and the rack using grooves and guide rails, a guide pin mechanism may be provided on the rack and chassis, which may further align the rack and chassis. The guide pin mechanism may include pins on the rack and mating holes on the chassis. After the initial bite, the robot can again move down faster, until it reaches the second alignment mechanism (or several inches further than the size of the guide pin, for example 2 inches further). Here, the second alignment mechanism may be a guide pin mechanism. The robot moves slowly downward so that the pins on the rack can connect with the occlusal holes on the chassis. The robot slowly continues to move downward until the load cell provides feedback indicating that the chassis has been inserted, eg, the load cell detects positive pressure. At this point, the linkage mechanism may be inactive and the robot may move upwards and back to its home position (a few inches slowly and then quickly to ensure proper placement).

In one exemplary embodiment, the chassis or rack may include presence detection pins. When the presence detection pin is mated with the corresponding receiver, the management system may receive a signal from the receiver. Such a signal may indicate that the chassis has been properly placed in position. In this exemplary embodiment, the robot may deactivate the association mechanism only after the receiver provides a signal to the management system.

During an ascending or descending operation, the management system may receive and monitor data received from the load cell as well as other sensors (eg, motion sensors, tile sensors, rotation sensors, accelerometers, etc.). Such data can ensure that the chassis does not collide or misalign, or that the robot has a good grip on the chassis. If the management system determines that the chassis has been somewhat bumped or misaligned, or that the robot has been tilted or rotated (eg, due to a poor connection between the robot and the chassis), the management system may stop the lifting or lowering operation. have.

20A and 20B show an example robotic system 1100 . FIG. 20A shows a robotic system 1100 , which may be a gantry robot that includes a plate 1110 . The gantry robot can move within the tank and use the plate 1110 to raise the computing device. FIG. 20B shows a plate 1110 , which may include a linkage mechanism 1111 and a guide pin 1112 . The linkage mechanism 1111 may include a plurality of fingers 1113 mechanically coupled to one or more armatures. Once the linkage mechanism is disposed within the connecting plate of the chassis, the armature can be activated and the fingers can move to hold the connecting plate.

21A and 21B show an exemplary guide pin mechanism between the chassis 1150 and the rack 1160 . In this exemplary embodiment, the rack 1160 may include two guide pins 1161 (for each chassis 1150 ), the chassis 1150 having two guide pins 1161 configured to receive them. It may include two occlusal holes 1151 . When the robot lowers the chassis 1150 onto the rack 1160 , the guide pin mechanism ensures proper electrical connection between the chassis 1150 and the rack 1160 .

In one exemplary embodiment, the robot is a robotic arm. The robot arm can move on rails provided on the side of the tank. In one exemplary embodiment, each chassis may be pulled using a piston and connected to a channel located at the upper end of the piston. The channel may convey the chassis to the magazine, for example using a rail system.

In one exemplary embodiment, the robot may include a calibration system that may include a plurality of sensors. The calibration system may determine whether the robot has exceeded its normal operating range. For example, a tilt sensor may notify the robot that the robot is out of balance or tilted. As another example, a load cell may provide a signal to the robot when the robot is not moving freely, for example, when it collides with an object.

In one exemplary embodiment, the robot may use artificial intelligence or machine learning techniques to provide hot swapping or as a failsafe mechanism.

In one exemplary embodiment, the container may include a plurality of cameras. In this example, one camera may be mounted to the robot and the other may be mounted to the container wall. The cameras may be mounted in such a way that the user always has a view of the moving component of the container. The container may also include a user interface displayed on a display device, such as a monitor. As the robot raises or lowers the chassis, the user interface may display video fed from the camera. In this way, if something goes wrong with the robot's motion, the user can take action.

In one exemplary embodiment, the robotic system may be a vision-based system coupled to active control. Active control allows the reference point to be passed back through the logic, which will determine the proximity to proximity switching. In one exemplary embodiment, the robotic system may be an AI robotic system. In one exemplary embodiment, the robotic system may be an automatic calibration system. In one embodiment, the robotic system may be a logic controlled active loop based system that is preprogrammed and capable of calibrating based on the distance from the idle state to the active state.

Absorption/desorption unit

In one exemplary embodiment, the absorption unit may be a carbon-bed-based system. The absorption unit may be a round circular drum. An aluminum framework may be positioned inside the absorbent unit, which may include copper ribbon heating elements spread throughout the framework. The height and radius of the absorption unit can be designed based on the size of the container and the volume of fluid in the tank.

The absorbent unit may be sealed and may contain activated carbon within the framework. The absorption unit may include an inlet and an outlet. In one embodiment, the absorption unit may include a cooling system, eg, cold air may flow through the center of the absorption unit without contacting the carbon. Such a system makes it possible to cool the carbon through convection.

In one exemplary embodiment, the activated carbon enables the absorption or attachment of dielectric vapors. When balancing the tank (eg, creation of pressure or vacuum) is required, the management system can connect the absorption unit to the tank by opening a valve. The management system may activate or initiate power to the copper heating ribbon element, which may heat the carbon. The carbon then releases the dielectric fluid molecules as vapor, which can be returned back to the tank.

In one exemplary embodiment, pressure and temperature sensors may be located within the carbon bed to prevent pressure or temperature exceeding conditions.

In one exemplary embodiment, the absorption unit may include a control loop bypass to relieve (or pressurize) the pressure in an emergency. This is a safety feature of the container. The absorption unit has a valve that separates the absorption unit from the tank. In the event of a valve failure, an overpressure condition may exist. For example, if the outlet of the absorption unit is blocked, the management system may open a bypass valve. When the bypass valve is opened, the dielectric vapor can proceed to the atmosphere, thereby preventing explosion of the absorption unit. When there is a vacuum condition in the tank, the valve can be opened so that air can flow into the tank and prevent collapsing of the tank.

In one exemplary embodiment, the vessel may include a plurality of safety bypass valves. For example, during a startup operation, the computing device may generate excessive amounts of steam. The valve that allows steam to escape from the tank to the absorption unit may fail. Thus, there may be an overpressurization condition in the tank that the bellows cannot handle. An emergency bypass valve may be opened to release some of the steam to the atmosphere.

As another example, during a start-up operation, an excess of vapor may enter the absorption unit. This can create over-pressurized conditions within the absorption unit. Thus, the bypass valve of the absorption unit can be opened to release the vapor to the atmosphere.

In one exemplary embodiment, in addition to pressure and temperature, the management system may receive and monitor data from the absorption unit related to the absorption unit's power. The management system ensures that current flows through the absorption unit. The management system can shut off the absorption unit if there is an overcurrent problem or an over-pressurized condition.

Self-alignment of chassis

In one exemplary embodiment, the chassis may include self-aligning features. The self-aligning feature may include a plate that may be moved (ie, floated) relative to the chassis. There may be one or more input ports or output ports (or connectors) on the plate. The chassis (and plate) may include an occlusal hole that may receive a guide pin to align the plate to receive ports. In one exemplary embodiment, the port may be an ExaMAX® connector.

In one exemplary embodiment, the self-aligning feature may include several alignment mechanisms. For example, as a first self-aligning mechanism, the rack and chassis may have grooves and guide rails. As a second self-aligning mechanism, the rack may include tapered, rounded-ended pins. The pins may run into catch holes in the chassis. When the pin is inserted into the mating hole (or catch hole), it results in a final precise alignment between the connector on the chassis and the interface on the rack (ie the backplane). By the time the connector is ready to mate with its occlusal portion, the alignment pins ensure that the floating pair is perfectly aligned with the relative orientation of the connector being approached for insertion.

In one exemplary embodiment, the connectors of the chassis and rack may include their own alignment mechanism, eg pins may be part of the connector.

In one exemplary embodiment, the connector may include a multi-stage mechanism for self-alignment, including a full outer alignment catch and a finer inner alignment catch.

In one exemplary embodiment, the plate may be the backplane of the rack.

22 depicts an exemplary connector with self-aligning features. Such connectors may include guiding pins and other guiding features to ensure proper connection between the pairs.

The process of immersion cooling

1. The method is:

At least partially immersing a computer component in a thermally conductive condensable dielectric fluid comprising:

The computer component is mounted within a chassis including a backplane for receiving power from the rack; and

wherein the computer component is configured to dissipate heat within the dielectric fluid when the computer component is operated;

using a condenser to condense the gas phase of the dielectric fluid to the liquid phase of the dielectric fluid;

wherein the rack is positioned within a tank comprising a pressure controller for decreasing or increasing the pressure inside the tank.

2. The method of embodiment 1, wherein the tank has a computing power density of at least 300 W of computing power distributed over each square foot of space.

3. The method of embodiment 1, further comprising removing the chassis from the rack using a robot, wherein the robot is positioned within the tank.

4. The method according to embodiment 3, further comprising, using the robot, transferring the chassis to an airlock, wherein the airlock is disposed on the inner side of the tank without substantially impeding pressure in the tank. A method configured to allow access to.

5. The method of Example 4,

opening an inner door of the airlock;

placing the chassis within the airlock;

closing the inner door of the airlock;

making the pressure of the airlock equal to atmospheric pressure; and

and opening an outer door of the airlock.

6. The method according to embodiment 3, further comprising using the robot to store the chassis in a magazine.

7. The method of embodiment 6, wherein the magazine is disposed on a platform comprising a support member, a rotating member, and a rail.

8. The method of embodiment 3, wherein the robot is a gantry robot configured to remove, replace, or install the chassis.

9. The method according to embodiment 8, wherein the gantry robot is configured to move on a horizontal plane and to be lowered vertically.

10. The apparatus of embodiment 9, wherein the robot is configured to remove, replace, or install a component of a power distribution system.

11. The method of embodiment 10, wherein the robot comprises a gripping tool for gripping the chassis.

12. The method of embodiment 1, wherein the tank is mounted within a superstructure comprising a plurality of tanks.

13. The method of embodiment 1, further comprising removing contaminants from the dielectric fluid.

14. The method of embodiment 1, further comprising removing gaseous contaminants.

15. The method of embodiment 1, further comprising providing power, a network connection, and a process fluid to the tank.

16. The method of embodiment 1, wherein the tank comprises an opening in a top end and a removable cover.

17. The method of embodiment 1, wherein the tank comprises an inner volume of from about 100 cubic feet to about 300 cubic feet.

18. The method of embodiment 1, wherein the chassis does not include a heat sink and a fan.

19. The method of embodiment 1, wherein the chassis comprises a blade server, a processor, a power source, or an interface card.

20. The apparatus of embodiment 19, wherein the backplane is electrically coupled to the interface card, which is Cat6A or Cat7 compatible with an RJ45 interface for connection to a 1G or 10G Ethernet interface.

Vessel design and construction for immersion cooling

1. The device is:

a tank configured to hold a thermally conductive condensable dielectric fluid;

a pressure controller for decreasing or increasing the pressure inside the tank;

a rack at least partially submerged in the dielectric fluid;

a condenser for condensing the gas phase of the dielectric fluid; and

and a robot configured to move a chassis within the rack.

2. The apparatus of embodiment 1, wherein the apparatus comprises a modular skid comprising a plurality of tubes for a forklift.

3. The apparatus of embodiment 1, wherein the tank has a computing power density of at least 300 W of computing power distributed over each square foot of space.

4. The apparatus of embodiment 1, wherein an outer portion of the apparatus includes a power input and a communication input.

5. The method of Example 4,

the power input and the communication input are electrically connected to the box; and

wherein the box distributes power input and communication input to the rack using a plurality of wires.

6. The apparatus of embodiment 5, wherein the rack comprises a backplane receiver configured to distribute power and communication signals to the chassis.

7. The method of embodiment 6, wherein the chassis comprises a backplane, the backplane comprising:

to receive power and communication signals from a backplane receiver of the rack; and

and distribute the power and communication signals to computer components within the chassis.

8. The apparatus of embodiment 5, wherein the plurality of wires do not include plastic insulation.

9. The apparatus of embodiment 5, wherein the rack comprises a transformer.

10. The device of embodiment 1, wherein the device is stackable.

11. The apparatus of embodiment 1, wherein the apparatus comprises a magazine for storage of replacement components.

12. The apparatus of embodiment 11, wherein the robot is configured to remove the chassis from the rack and to place the chassis within a magazine.

13. The apparatus of embodiment 12, wherein the magazine is positioned on a platform comprising a rotating member, a support member, and a rail.

14. The apparatus of embodiment 13, wherein the platform is configured to guide the magazine out of the apparatus.

15. The apparatus of embodiment 1, wherein the apparatus comprises a dryer configured to remove water vapor contaminants from the apparatus.

16. The method of Example 1,

sump area;

Pump; and

further comprising a filter;

wherein the pump is configured to remove the dielectric fluid from the sump region and to pass the dielectric fluid through the filter prior to delivering the dielectric fluid to the bath portion of the tank.

17. The apparatus of embodiment 1, wherein the dielectric fluid has a boiling point in the range of 20°C to 100°C.

18. The device of example 1, wherein the dielectric fluid comprises a chemical of formula (CF3)2CFCF2OCH3, C4F9OCH3, CF3CF2CF2CF2OCH3, hydrofluoro ether or methoxy-nonaflurobutane.

19. The device of embodiment 1, further comprising a lock that disables the device from operating if any of the lid or door of the device is not secured.

20. The apparatus of embodiment 1, further comprising a controller configured to disconnect power to the apparatus in case of unauthorized access to the lid or door.

Robotics and automation for submerged cooling

1. The device is:

a tank configured to hold a thermally conductive condensable dielectric fluid;

a pressure controller for decreasing or increasing the pressure inside the tank;

a computer component at least partially immersed in the dielectric fluid;

a condenser for condensing the gas phase of the dielectric fluid; and

and a robot configured to pick up the computer component.

2. The apparatus of embodiment 1, further comprising an airlock.

3. The apparatus of embodiment 2, wherein the airlock comprises an inner door and an outer door.

4. The apparatus of embodiment 3, wherein the airlock is configured to receive an inert gas to purge the gas phase of the dielectric fluid before the outer door is opened.

5. The apparatus of embodiment 3, wherein the robot is located outside of the tank.

6. The apparatus of embodiment 3, wherein the robot is positioned within the tank.

7. The apparatus of embodiment 6, wherein the robot is configured to remove the computer component from the rack and deliver the computer component to the airlock.

8. The robot of embodiment 7, wherein:

to open the inner door of the airlock;

to place the computer component within the airlock;

to close the inner door of the airlock;

so that the pressure of the airlock is equal to atmospheric pressure; and

and open an outer door of the airlock.

9. The apparatus of embodiment 8, further comprising a second robot positioned outside the tank.

10. The apparatus of embodiment 9, wherein the second robot is configured to remove the computer component from the airlock when the outer door is opened.

11. The apparatus of embodiment 9, wherein the second robot is configured to place the computer component within a storage slot.

12. The apparatus of embodiment 9, wherein the airlock is configured to equalize the pressure of the airlock after the outer door is closed.

13. The apparatus of embodiment 1, wherein the apparatus is configured to receive instructions from a server located external to the apparatus.

14. The apparatus of embodiment 1, wherein the computer component is located within a chassis displaying an asset tag.

15. The apparatus of embodiment 14, wherein the robot is configured to scan the charity tag and relay the asset tag to a management system.

16. The apparatus of embodiment 1, wherein the robot is a gantry robot configured to remove, replace, or install the computer component.

17. The apparatus of embodiment 16, wherein the gantry robot is configured to move horizontally and vertically.

18. The apparatus of embodiment 1, wherein the robot is configured to remove, replace, or install a component of a power distribution system.

19. The apparatus of embodiment 18, wherein a component of the power distribution system is a transformer or a power source.

20. The apparatus of embodiment 1, wherein the robot comprises a gripping tool for gripping the computer component.

Ballast block for submerged cooling

1. The device is:

As a tank:

a bath portion for holding thermally conductive condensable dielectric fluid and computer components; and

a tank comprising a shelf portion configured to hold at least one ballast block;

a pressure controller for decreasing or increasing the pressure inside the tank;

a condenser for condensing the gas phase of the dielectric fluid; and

and a robot configured to pick up the computer component.

2. The apparatus of embodiment 1, wherein the bottom point of the bath portion has a height lower than the height of the shelf portion.

3. The apparatus of embodiment 1, wherein the bath portion is configured such that the computer component is at least partially immersed in the dielectric fluid.

4. The apparatus of embodiment 3, wherein the computer component is a blade server, processor, power source, or transformer.

5. The apparatus of embodiment 1, wherein the level of dielectric fluid is high enough to cover at least a portion of the shelf portion.

6. The apparatus of embodiment 1, wherein the shelf portion is located next to the condenser.

7. The apparatus of embodiment 6, wherein the shelf portion is configured to receive condensed dielectric fluid from the condenser.

8. The apparatus of embodiment 1, wherein the ballast block is configured to occupy a volume of a tank above the shelf for displacing the dielectric fluid from the shelf to an area above the bath portion.

9. The apparatus of embodiment 1, wherein the ballast block comprises a plurality of riser pits for allowing the dielectric fluid to flow under the ballast block.

10. The apparatus of embodiment 1, wherein the ballast block is insoluble in the dielectric fluid.

11. The apparatus of embodiment 1, wherein the ballast block is made of metal, rubber, silicone, or polymer.

12. The apparatus of embodiment 1, wherein the ballast block is denser than the dielectric fluid.

13. The apparatus of embodiment 1, wherein the ballast block has a handle, cutout or plate for removal or replacement of the ballast block.

14. The apparatus of embodiment 13, wherein the robot is configured to raise the ballast block using the handle, cutout or plate.

15. The apparatus of embodiment 1, wherein the ballast block is configured to interlock with another ballast block from a top side or a bottom side of the ballast block.

16. The apparatus of embodiment 15, wherein the interlocking prevents the other ballast block from sliding.

17. The apparatus of embodiment 15, wherein the other ballast block is configured to be positioned on a top side or a bottom side of the ballast block.

18. The ballast block of embodiment 15, wherein the ballast block comprises a recessed portion on the top side of the ballast block, such that the raised pit of the other ballast block is configured to lock within one of the recessed portions of the ballast block. becoming a device.

19. The apparatus of embodiment 1, wherein the ballast block is configured to span at least 40% of the total length of the shelf portion.

20. The apparatus of embodiment 1, wherein the ballast block has external dimensions of about 2 feet long, about 8 inches wide, and about 1 inch high.

Server case for submerged cooling

1. The device is:

a tank configured to hold a thermally conductive condensable dielectric fluid;

a pressure controller for decreasing or increasing the pressure inside the tank;

a chassis at least partially submerged in the dielectric fluid;

a condenser for condensing the gas phase of the dielectric fluid; and

and a robot configured to pick up the chassis.

2. The apparatus of embodiment 1, wherein the chassis does not require a heat sink and fan.

3. The apparatus of embodiment 1, wherein the chassis comprises a blade server.

4. The apparatus of embodiment 1, wherein the chassis comprises a processor, a power supply, or an interface card.

5. The apparatus of embodiment 4, wherein the interface card is Cat6A or Cat7 compatible with an RJ45 interface for connection to a 1G or 10G Ethernet interface.

6. The apparatus of embodiment 1, wherein the chassis is removably attached to a rack.

7. The apparatus of embodiment 6, wherein the chassis comprises a backplane for providing a slot-in interface between the chassis and the rack.

8. The apparatus of embodiment 7, wherein the backplane is configured to distribute power and signals received from the rack within the chassis.

9. The apparatus of embodiment 8, wherein the backplane is configured to transmit power and data via a cable to the blade server.

10. The chassis of embodiment 1, wherein the chassis is a substantially rectangular box comprising a rear wall and two side walls, the rear wall having a plurality of holes for assisting circulation of the dielectric fluid within the chassis. Device.

11. The apparatus of embodiment 10, wherein the chassis comprises guide rails at each of the two sidewalls.

12. The apparatus of embodiment 1, wherein the chassis comprises a mounting interface for holding computer components.

13. The apparatus of embodiment 1, wherein the chassis comprises a plate, and wherein the robot is configured to raise the chassis using the plate.

14. The apparatus of embodiment 1, wherein the chassis comprises a microcontroller.

15. The microcontroller of embodiment 14 wherein:

to receive, from a sensor mounted on the chassis, sensor data indicating whether the chassis is properly positioned within a rack; and

and send the sensor data to a management system.

16. The microcontroller of embodiment 14 wherein:

to receive a power signal from the management system; and

and send the power signal to a switch configured to cut off power in the chassis.

17. The microcontroller of embodiment 14 wherein:

to receive operational data from a computer component mounted within the chassis; and

and send the motion data to a management system.

18. The apparatus of embodiment 14, wherein the microcontroller is configured to control electrical and communication facilities of the blade server.

19. The apparatus of embodiment 1, wherein the chassis comprises an RFID tag.

20. The apparatus of embodiment 19, wherein the robot is configured to scan an RFID tag and send a signal to a management system.

Steam management for submerged cooling with bellows

1. The device is:

a tank configured to hold a thermally conductive condensable dielectric fluid and a computer component;

a pressure controller for decreasing or increasing the pressure inside the tank;

a vapor management system for condensing the gas phase of the dielectric fluid; and

and a robot configured to pick up the computer component.

2. The apparatus of embodiment 1, wherein the vapor management system comprises a condensing structure within the tank.

3. The apparatus of embodiment 2, wherein the condensing structure comprises thermally conductive tubes, coils or radiator fins.

4. The apparatus of embodiment 2, wherein the condensing structure is configured to be coupled to a source of cooling liquid such that the cooling liquid passes through the condensing structure.

5. The apparatus of embodiment 2, wherein the apparatus is configured to quench the cooling liquid using an evaporative cooling or dry cooling tower.

6. The apparatus of embodiment 2, wherein the vapor management system comprises an inlet pipe and an outlet pipe.

7. The apparatus of embodiment 6, wherein the inlet pipe is configured to receive cooling liquid from a quench cooling liquid source and to guide cooling liquid through the condensing structure.

8. The apparatus of embodiment 6, wherein the outlet pipe is configured to receive cooling liquid from the condensing structure and return the cooling liquid to the quench cooling liquid source.

9. The apparatus of embodiment 1, wherein the vapor management system comprises a storage unit for storing the dielectric fluid.

10. The apparatus of embodiment 9, wherein the vapor management system is configured to direct dielectric fluid in the tank from the storage unit.

11. The apparatus of embodiment 1, wherein the vapor management system comprises a vapor storage unit for storing vapor of the dielectric fluid.

12. The apparatus of embodiment 11, wherein the vapor storage unit is a bellows.

13. The apparatus of embodiment 12, wherein the bellows is configured to expand or contract to maintain pressure inside the tank.

14. The device of embodiment 12, wherein the bellows comprises one or more pouches.

15. The apparatus of embodiment 11, wherein the vapor storage unit comprises a valve allowing air to flow into the vapor management system to lower the temperature of the vapor of the dielectric fluid.

16. The apparatus of embodiment 15, wherein the vapor storage unit is operatively connected to a carbon bed to separate vapor of the dielectric fluid from air.

17. The apparatus of embodiment 16, wherein the carbon bed comprises a desorption heater configured to heat the carbon bed to increase the temperature of the carbon bed.

18. The apparatus of embodiment 1, wherein the vapor management system comprises a filter.

19. The apparatus of embodiment 17, wherein the filter is configured to remove air and water vapor.

20. The vapor management system of embodiment 1 comprising:

an inert gas storage unit; and

and during a start-up operation or a shutdown operation, to introduce an inert gas from the inert gas storage unit into the tank.

Claims (15)

The system is:
a tank configured to hold a liquid phase and a gas phase of a fluid;
a structure in the tank configured to maintain one or more computer components at least partially submerged in the liquid phase of the fluid during operation of the system;
a heating element configured to heat the liquid phase of the fluid; and
and a controller configured to regulate the heating element. According to claim 1,
wherein the heating element is configured to be completely submerged in the liquid phase during operation of the system. According to claim 1,
wherein the controller is configured to regulate the heating element using a matrix. According to claim 1,
wherein the system further comprises a temperature sensor or a pressure sensor, the sensor operatively coupled to the controller. According to claim 1,
The controller is:
to receive data related to the operating load, temperature, or both of the processor; and
and adjust the heating element based on an operating load of the processor, a temperature, or both. According to claim 1,
wherein the controller is configured to cause the heating element to heat the liquid phase of the fluid during or prior to a start-up operation. According to claim 1,
wherein the controller is configured to cause the heating element to heat the liquid phase of the fluid when the temperature of the liquid phase of the fluid is below a threshold temperature. According to claim 1,
wherein the controller is configured to cause the heating element to heat the liquid phase of the fluid during or prior to a boost operation. 9. The method of claim 8,
wherein the boost operation is identified by the controller in response to an indication that an operating load of a processor is to be increased. According to claim 1,
further comprising a pressure management system, wherein the controller is configured to activate pressure management. According to claim 1,
wherein the controller is configured to operate the heating element to maintain the temperature of the liquid phase of the fluid within a threshold range below the boiling point of the fluid. A method of cooling computer components is:
Receiving first sensor data from a sensor located within a tank, the tank comprising:
(a) the liquid and gas phases of a fluid; and
(b) one or more computer components at least partially immersed in the liquid phase of the fluid;
(c) a heating element at least partially submerged in the liquid phase of the fluid; and
(d) configured to hold the sensor;
activating the heating element based on sensor data; and
deactivating the heating element based on sensor data. 13. The method of claim 12,
and detecting an operating mode of the one or more computer components, wherein the operating mode is a startup mode, a boost mode, a deceleration mode, or a halt operation. 14. The method of claim 13,
wherein the startup mode is when the one or more computer components are inactive for a period of time prior to operation of the one or more computer components. 14. The method of claim 13,
wherein the boost mode is when an increase in the workload of the one or more computer components is expected.

Images