TWI787673B - Hybrid cluster system and computing node thereof - Google Patents

Abstract

A hybrid cluster system includes at least one computing node for providing computing resources and at least one storage node for providing storage resources. A specification of the at least one computing node is identical to a specification of the at least one storage node.

Description

Hybrid Cluster System and Its Computing Nodes

The invention refers to a hybrid cluster system and its computing node, especially the hybrid cluster system and its computing node which are beneficial to system update and can improve product reusability and flexibility.

Most of the existing servers have special specifications, which are not compatible with the system interface of other servers, and there is no uniform specification and size. Therefore, the update or upgrade of the system can only rely on the original design manufacturer, which hinders the update or upgrade. In addition, the existing servers are usually only used as computing nodes and cannot support the integration of storage devices. If there is a storage requirement, an additional storage server needs to be configured. Therefore, how to save design costs and integrate storage and computing requirements has become an important issue.

Therefore, the present invention mainly provides a hybrid cluster system and its computing nodes, which is beneficial to system update and improves product reusability and flexibility.

The present invention discloses a hybrid cluster system, comprising at least one storage node for providing storage resources; and at least one computing node for providing computing resources, the specification of the at least one computing node is the same as the specification of the at least one storage node .

The present invention also discloses a computing node for providing computing resources, including a plurality of computing elements, the computing node is coupled to a storage node, and the specification of the computing node is the same as that of the storage node.

10,20,70: Hybrid cluster system

Nsoc1, Nsoc2, Nsoc3, Nsoc7: computing nodes

Nhdd1, Nhdd2: storage nodes

210: Chassis

220: Backplane

230,530: Exchanger

313: random access memory

315: flash memory

317: Computing element

319,532,534,622,629: Connectors

538: Management chip

Px86:x86 platform server

SR1~SR5: User

FIG. 1 is a schematic diagram of a hybrid cluster system in an embodiment of the present invention.

FIG. 2A is a schematic diagram of a hybrid cluster system according to an embodiment of the present invention.

FIG. 2B is a schematic diagram of the hybrid cluster system shown in FIG. 2A.

FIG. 3 is a schematic diagram of a computing node in an embodiment of the present invention.

FIG. 4 is a schematic diagram of component configuration of the computing node shown in FIG. 3 .

Fig. 5 is a schematic diagram of a switch in an embodiment of the present invention.

FIG. 6 is a schematic diagram of a backplane in an embodiment of the present invention.

FIG. 7 is a schematic diagram of a hybrid cluster system, an x86 platform server and users in an embodiment of the present invention.

"Include" mentioned in the entire specification and subsequent claims is an open term, so it should be interpreted as "including but not limited to". The descriptions such as "first" and "second" mentioned in the entire specification and subsequent claims are only used to distinguish different components, and do not limit the order of their production.

Please refer to FIG. 1 , which is a schematic diagram of a hybrid cluster system 10 in an embodiment of the present invention. The hybrid cluster system 10 may include a computing node Nsoc1 and a storage node Nhdd1, accordingly, the hybrid cluster system 10 can be used to provide computing and storage resources to integrate storage and computing requirements. The computing node Nsoc1 is used to virtualize the user's virtual platform, and the computing node Nsoc1 may be an Advanced RISC Machine (ARM) micro server, but not limited thereto. The storage node Nhdd1 is used to store data, and the storage node Nhdd1 may be a 2.5-inch Hard Disk Drive (2.5-inch HDD), but not limited thereto. The size of the computing node Nsoc1 is the same as that of the storage node Nhdd1, for example, both adopt the standard specification of 2.5 inches in the prior art. Moreover, the interface of the computing node Nsoc1 is the same as the interface of the storage node Nhdd1. In some embodiments, both the computing node Nsoc1 and the storage node Nhdd1 use SFF-8639 connectors. In some embodiments, both the computing node Nsoc1 and the storage node Nhdd1 use a Non-Volatile Memory Express (NVMe) interface. In some embodiments, both the computing node Nsoc1 and the storage node Nhdd1 use a Peripheral Component Interconnect Express (PCIe) interface. In some embodiments, the interface of the computing node Nsoc1 and the interface of the storage node Nhdd1 both support hot swapping/hot plugging.

In short, the specification of the computing node Nsoc1 is the same as that of the storage node Nhdd1. In this way, the computing node Nsoc1 can be compatible with the system interface set by the storage node Nhdd1, thereby saving design costs and improving product reusability. Moreover, the computing node Nsoc1 and the storage node Nhdd1 can be replaced with each other, for example, the storage node Nhdd1 originally set can be changed to the computing node Nsoc1, which is beneficial to system upgrade or update. Furthermore, the configuration ratio of the number of computing nodes Nsoc1 and the number of storage nodes Nhdd1 can be changed and adjusted according to different requirements, thereby increasing product flexibility.

Specifically, please refer to FIG. 2 . FIG. 2A is a schematic diagram of a hybrid cluster system 20 according to an embodiment of the present invention, and FIG. 2B is a schematic diagram of the hybrid cluster system 20 shown in FIG. 2A . The hybrid cluster system 20 can be used as the hybrid cluster system 10 . The hybrid cluster system 20 may include a chassis 210, a rear A backplane board 220, a switch 230, a computing node Nsoc2, and a storage node Nhdd2. The chassis 210 is used for accommodating the backplane 220 , the switch 230 , the computing node Nsoc2 and the storage node Nhdd2 . The backplane 220 is electrically connected between the switch 230, the computing node Nsoc2, and the storage node Nhdd2, so that the computing node Nsoc2 can be coupled to the storage node Nhdd2. A backplane 220 may include a plurality of bays arranged in an array, and the bays have a fixed interval. The computing node Nsoc2 or the storage node Nhdd2 is respectively plugged into the slot of the backplane 220 to be electrically connected to the backplane 220. In this way, the backplane 220 can perform power transmission and signal transmission with the computing node Nsoc2 or the storage node Nhdd2. On the other hand, the switch 230 can address the computing node Nsoc1 and the storage node Nhdd1 of the hybrid cluster system 20 .

The computing node Nsoc2 and the storage node Nhdd2 can serve as the computing node Nsoc1 and the storage node Nhdd1 respectively. In some embodiments, the storage node Nhdd2 may be a non-volatile memory (Non-Volatile Memory), but not limited thereto. In some embodiments, data can be distributed and stored in different storage nodes Nhdd2. The storage node Nhdd2 may be disposed in a chassis, and the size of the chassis is the size of the storage node Nhdd2. In some embodiments, the size of the computing node Nsoc2 may be smaller than or equal to the size of the storage node Nhdd2. In some embodiments, in some embodiments, both the computing node Nsoc1 and the storage node Nhdd1 conform to the 2.5-inch hard disk form (Form Factor), but not limited thereto, the computing node Nsoc1 and the storage node Nhdd1 can also conform to 1.8-inch hard drive form or 3.5-inch hard drive form. In some embodiments, the interface of the computing node Nsoc1 and the interface of the storage node Nhdd1 are the same, for example, both adopt the standard SFF-8639 non-volatile memory host controller interface specification interface. Since the computing node Nsoc2 and the storage node Nhdd2 have the same size and interface, the computing node Nsoc2 can be compatible with the system interface (for example, the system interface used in the prior art) set by the storage node Nhdd2, that is, the shared chassis 210 (For example, it may be a chassis used in the prior art), so as to save design costs and improve product reusability.

In addition, since the computing node Nsoc2 can be accommodated in the slot of the storage node Nhdd2, the configuration ratio of the number of computing nodes Nsoc2 and the number of storage nodes Nhdd2 can be changed and adjusted according to different requirements. For example, in some embodiments, the hybrid cluster system 20 may include 3 backplanes 220 , and one backplane 220 may include 8 slots, but not limited thereto. That is to say, the hybrid cluster system 20 may include 24 slots for the computing node Nsoc2 and the storage node Nhdd2 to be plugged into the backplane 220, and the total number of the computing node Nsoc2 and the storage node Nhdd2 is limited to a certain value ( e.g. 24). As shown in FIG. 2, the hybrid cluster system 20 may include 20 computing nodes Nsoc2 and 4 storage nodes Nhdd2, but it is not limited thereto. The hybrid cluster system 20 may also include only 18 computing nodes Nsoc2 and 5 storage nodes Nhdd2, but not all the slots are filled. That is to say, the number ratio between the computing node Nsoc2 and the storage node Nhdd2 is adjustable. The 24 slots of the hybrid cluster system 20 can be evenly arranged at fixed intervals, so that the computing nodes Nsoc2 or storage nodes Nhdd2 plugged into the slots of the backplane 220 are aligned to four planes (that is, the chassis 210 The bottom surface and the top surface of the backplane 220 and the front panel opposite to the backplane 220) are arranged. As shown in Figure 2, 20 computing nodes Nsoc2 are set on the left side of the hybrid cluster system 20, and 4 storage nodes Nhdd2 are set on the right side of the hybrid cluster system 20, that is to say, the computing node Nsoc2 and the storage node Nhdd2 can be based on Category arrangement. However, the present invention is not limited thereto. As shown in FIG. 1 , computing nodes Nsoc1 and storage nodes Nhdd1 may also be arranged in a staggered manner.

Please refer to FIG. 3 , which is a schematic diagram of a computing node Nsoc3 in an embodiment of the present invention. Computing node Nsoc3 may serve as computing node Nsoc1. The computing node Nsoc3 may include a random access memory (Random Access Memory, RAM) 313 , a flash memory (Flash Memory) 315 , a computing element 317 and a connector (connector) 319 . The computing element 317 is coupled between the random access memory 313 , the flash memory 315 and the connector 319 . In some embodiments, the data communication link (data communication link) can conform to the peripheral component interconnection standard. In some embodiments, an operating system, such as a Linux operating system, can be stored in the random access memory 313 . In some embodiments, the computing element 317 can be a System on a Chip, which can process digital signals, analog signals, mixed signals or even higher frequency signals, and can be applied in embedded systems. In some embodiments, computing element 317 may be an RISC machine SoC. As shown in FIG. 3 , the computing node Nsoc3 includes two computing elements 317 , but not limited thereto, and the computing node Nsoc3 may include more than two computing elements 317 . The connector 319 supports power and signal transmission, and supports hot plugging. In some embodiments, connector 319 may employ a peripheral component interconnection standard interface. In some embodiments, connector 319 may be a SFF-8639 connector. SFF-8639 can be called U.2 interface, which is regulated by SSD Form Factor Work Group. FIG. 4 shows a schematic diagram of the component configuration of the computing node Nsoc3 shown in FIG. 3 . However, the component configuration of the computing node Nsoc3 is not limited to the configuration shown in FIG. 4 , and can be adjusted according to different design considerations.

Please refer to FIG. 5 , which is a schematic diagram of a switch 530 in an embodiment of the present invention. The switch 530 may serve as the switch 230 . The switch 530 can be an Ethernet switch or other switches. The switch 530 may include connectors 532 , 534 and a management chip 538 . The management chip 538 is coupled between the connectors 532 , 534 . The data communication link between the connectors 532 , 534 and the management chip 538 can conform to the Peripheral Device Interconnection standard. The connector 532 can be a board-to-board (Board to Board, B2B) connector, but not limited thereto. The connector 534 can be an SFP28 connector, but not limited thereto. Connector 534 can serve as a network interface. The switch 530 can route the data signal from the connector 534 to one of the computing elements of a plurality of computing nodes (such as the computing element 317 of the computing node Nsoc3 shown in FIG. 3 ). The management chip 538 can be a field programmable logic gate array (Field Programmable Gate Array, FPGA), but not limited thereto, and the management chip 538 can also be a programmable logic controller (programmable logic controller, PLC) or a special application integrated circuit (Application Specific Integrated Circuit, ASIC). In some embodiments, the management chip 538 can be used to manage computing nodes and storage nodes (such as computing node Nsoc2 and storage node Nhdd2 shown in FIG. 2 ). In some embodiments, the management chip 538 can be used to manage computing elements of a computing node (eg, computing element 317 of computing node Nsoc3 shown in FIG. 3 ).

Please refer to FIG. 6 , which is a schematic diagram of a backplane 620 in an embodiment of the present invention. The backplane 620 can serve as the backplane 220 . The backplane 620 may include connectors 622 , 629 . The data communication link between the connectors 622 and 629 can conform to the peripheral component interconnection standard. The connector 622 can be a board-to-board connector, but not limited thereto. The connector 629 supports power and signal transmission, and supports hot plugging. Connector 629 may be a SFF-8639 connector. The backplane 620 can be used to relay management data, so that the data is transmitted between the switch (such as the switch 230 shown in FIG. 2 ) and the corresponding computing node (such as the computing node Nsoc2 shown in FIG. 2 ). Since the hybrid cluster system (such as the hybrid cluster system 20 shown in FIG. 2 ) may not include a central processing unit (Central Processing Unit, CPU) and is different from the existing way of managing servers, the backplane 620 can also Contains a microprocessor (Microprocessor) to assist the management chip of the switch (such as the management chip 538 of the switch 530 shown in FIG. 5) to manage the computing elements of the computing node (such as the computing node Nsoc3 shown in FIG. 3 computing element 317).

Please refer to FIG. 7, which is a schematic diagram of a hybrid cluster system 70, an x86 platform server Px86, and users SR1-SR5 in an embodiment of the present invention. The hybrid cluster system 70 can serve as the hybrid cluster system 10 . In some embodiments, the hybrid cluster system 70 adopts the kernel of the Linux operating system. The hybrid cluster system 70 may comprise a plurality of computing nodes Nsoc7, and the number of computing nodes Nsoc7 of the hybrid cluster system 70 may be appropriately adjusted according to different models. For example, the hybrid cluster system 70 may contain more than 30 Compute node Nsoc7. The computing node Nsoc7 in the hybrid cluster system 70 may be a micro-server of an RISC platform. Compared to x86 platform servo The processor Px86 and the computing node Nsoc7 of the hybrid cluster system 70 have high cost performance, that is to say, the cost and power consumption are lower under the same performance. The hybrid cluster system 70 connects micro-servers (i.e. computing nodes Nsoc7) with advanced RISC architecture in series to form a huge computing center. In this way, the operating performance of the application program (Mobile Application, APP) can be improved, reducing cost and power consumption.

Specifically, a computing node Nsoc7 is virtualized into a hybrid cluster system 70 of multiple mobile devices (such as mobile phones) through virtual technology, which can provide cloud services for mobile application (Mobile Application, APP) streaming (Streaming) platforms . Users SR1~SR5 do not need to download various applications, and can directly connect to the cloud to run all required applications (such as mobile games, group control marketing), and move the computing load to the data center for processing. That is to say, all calculations are completed in the data center, and the images or sounds generated by the devices of users SR1~SR5 are all processed in the data center before being streamed to the devices of users SR1~SR5. Since mobile devices are built in the hybrid cluster system 70 in a virtualized manner, users SR1~SR5 only need to connect to the x86 platform server Px86 through the Internet to log in to their accounts, and then users SR1~SR5 can remotely access their own devices. The terminal (remote) operates the virtual mobile device in the hybrid cluster system 70, and runs all required applications (such as mobile games, group control marketing), without downloading and installing applications to the devices of users SR1~SR5, so there will be no It is limited by the hardware specifications of the devices of users SR1~SR5. In this way, users SR1-SR5 can reduce the risk of device poisoning, save device space and improve operation performance. Program developers can save maintenance costs (such as information security maintenance) to ensure that applications can run on various devices. Furthermore, in some embodiments, the computing node Nsoc7 of the hybrid cluster system 70 can be used to store the resource files (such as program code, library or environment configuration file) required by the Android (Android) application in the running The container is isolated from the outside (such as the Linux operating system) according to the sandbox (Sandbox) mechanism, so that changing the content of the running container will not affect the outside (such as the Linux operating system).

Since the hybrid cluster system 70 includes a computing node Nsoc7 and a storage node (such as the storage node Nhdd2 shown in FIG. 2 ), the hybrid cluster system 70 can perform computing and storage, and provide computing and storage resources. In some embodiments, one computing element (such as the computing element 317 shown in FIG. 3 ) can install a virtual platform, and one computing element can simulate 2 to 3 virtual mobile devices, but not limited thereto. In some embodiments, the computing element (such as the computing element 317 shown in FIG. 3 ) of the computing node Nsoc7 of the hybrid cluster system 70 provides an image processing function and supports image compression. In some embodiments, after the user SR1 logs in to the x86 platform server Px86, the x86 platform server Px86 assigns a virtual mobile device of a computing node Nsoc7 of the hybrid cluster system 70 to the user SR1, and the relevant information of the user SR1 (such as an application program) can be stored in a storage node of the hybrid cluster system 70 (such as the storage node Nhdd2 shown in FIG. 2 ). After the calculation node Nsoc7 completes the relevant calculation, it performs encoding (encode) to compress the image, and transmits it to the device of the user SR1 through the network. After the user SR1 receives the image, the device of the user SR1 performs decoding (decode) to generate the image. In this way, image traffic can be reduced, thereby achieving image acceleration.

To sum up, the computing nodes and storage nodes of the hybrid cluster system have the same specifications. In this way, the computing nodes can be compatible with the system interface set by the storage nodes, thus saving design costs and improving product reusability. Moreover, computing nodes and storage nodes can be replaced with each other, which is beneficial to system upgrade or update. Furthermore, the configuration ratio of the number of computing nodes and the number of storage nodes can be changed and adjusted according to different needs, thereby increasing product flexibility.

The above descriptions are only preferred embodiments of the present invention, and all equivalent changes and modifications made according to the scope of the patent application of the present invention shall fall within the scope of the present invention.

10: Hybrid cluster system

Nsoc1: computing node

Nhdd1: storage node

Claims (8)

A hybrid cluster system, comprising: at least one storage node for providing storage resources; at least one computing node for providing computing resources, the specification of the at least one computing node is the same as the specification of the at least one storage node; and a The backplane includes a plurality of slots, the plurality of slots are arranged in an array, the plurality of slots have a fixed pitch, the at least one computing node and the at least one storage node are respectively plugged into the plurality of slots of the backplane The slots are electrically connected to the backplane, and the backplane performs power transmission and signal transmission with the at least one computing node. The hybrid cluster system as claimed in claim 1, wherein the at least one computing node and the at least one storage node are in the form of 2.5-inch hard drives. The hybrid cluster system as claimed in claim 1, wherein the at least one computing node and the at least one storage node both use a non-volatile memory host controller interface standard interface. The hybrid cluster system as claimed in claim 1, wherein a first connector of each of the at least one computing node and a second connector of each of the at least one storage node are SFF-8639 Connector. The hybrid cluster system as described in Claim 1, wherein the upper limit of the total number of the at least one computing node and the at least one storage node is a certain value, and the ratio of the number of the at least one computing node to the at least one storage node is configurable adjusted. The hybrid cluster system as claimed in claim 1, wherein the at least one computing node includes a plurality of computing elements, each of the plurality of computing elements is an RISC machine SoC, and the at least one computing node Each of the nodes is an RISC microserver. The hybrid cluster system as described in claim 1 further includes: a switch, the switch is an Ethernet switch, and the switch includes a network interface, and the switch is used to transfer data from the network interface The data signal of is routed to one of the at least one computing node. The hybrid cluster system as described in claim 1, wherein the at least one computing node and the at least one storage node are aligned to four planes, and the at least one computing node and the at least one storage node are arranged in a staggered manner or according to categories.

Images