CN206805432U - A kind of deep learning system based on rack-mount server - Google Patents
A kind of deep learning system based on rack-mount server Download PDFInfo
- Publication number
- CN206805432U CN206805432U CN201720056166.4U CN201720056166U CN206805432U CN 206805432 U CN206805432 U CN 206805432U CN 201720056166 U CN201720056166 U CN 201720056166U CN 206805432 U CN206805432 U CN 206805432U
- Authority
- CN
- China
- Prior art keywords
- water
- cooling
- rack
- deep learning
- system based
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
Landscapes
- Cooling Or The Like Of Electrical Apparatus (AREA)
Abstract
The utility model discloses a kind of deep learning system based on rack-mount server, including server unit, also include the water-cooling joint matched with server unit, the water intake end of the water-cooling joint connects one end of cold row by water pump, the water side of the water-cooling joint connects the other end of cold row, and fan is additionally provided with the cold row.Rack-mount server is radiated by way of water cooling, in the case where not being extended to space, the efficiency far of water-cooling is higher than wind-cooling heat dissipating.The utility model has the advantages that：Water-cooled cooling, cooling effectiveness are high；Intelligent control, the stability of a system is high, and robustness is good.
Description
Technical field
It the utility model is related to a kind of cooling system, and in particular to a kind of deep learning system based on rack-mount server
System.
Background technology
For information service enterprises, select to first have to consider volume, power consumption, caloric value of server etc. during server
Physical parameter, because information service enterprises are usually using large-scale Special machine room unified plan and manage substantial amounts of server resource,
Computer room is typically provided with tight security measures, good cooling system, the electric power system of multiple duplication, and the cost of its computer room is suitable
It is expensive.How the cost of serving of enterprise is directly connected in the limited more servers of interior volume administration, generally from machinery
Size meets the rack-mount server of 19 inch industrial standards.Rack-mount server also has a plurality of specifications, for example, 1U, 2U, 4U,
6U, 8U etc..Usual 1U rack-mount server most saves space, but performance and scalability are poor, is adapted to some business relative
Fixed uses field.More than 4U properties of product are higher, and scalability is good, the general high-performance processor for supporting more than 4
With substantial amounts of standard hot-plug part.Management is also very convenient, and the commonly provided people of manufacturer manages accordingly and monitoring tools, is adapted to
The crucial application of big visit capacity, but volume is larger, and space availability ratio is not high.
Rack-mount server will save space relative to tower server, but because spaces compact, radiating somewhat poor one
Point, simultaneously as deep learning needs computer to carry out the operation of long-time high capacity, radiating is again to stand in the breach to need to consider
The problem of, meanwhile, most of computer room spatial spreads are not easy, it is difficult to change spatial arrangement and air channel.
Utility model content
Technical problem to be solved in the utility model is that rack-mount server will save space relative to tower server,
But because spaces compact, radiate somewhat almost, simultaneously as deep learning needs computer to carry out long-time high capacity
Operation, radiating again be stand in the breach need consider the problem of, meanwhile, most of computer room spatial spreads are not easy, it is difficult to change space
Arrangement and air channel solve or else to carry out space, and it is an object of the present invention to provide a kind of deep learning system based on rack-mount server
The problem of radiating efficiency is improved in the case of extension.
The utility model is achieved through the following technical solutions：
A kind of deep learning system based on rack-mount server, including server unit, in addition to and server unit
The water-cooling joint of matching, the water intake end of the water-cooling joint connect one end of cold row, the water outlet of the water-cooling joint by water pump
End connects the other end of cold row, and fan is additionally provided with the cold row.Rack-mount server is dissipated by way of water cooling
Heat, in the case where not being extended to space, the efficiency far of water-cooling is higher than wind-cooling heat dissipating.
Temperature sensor is additionally provided with the water side of the water-cooling joint, is also associated with handling on the temperature sensor
Device, processor are also connected with water pump；
Temperature sensor：The temperature of water side cooling water is detected, temperature information is sent to processor；
Processor：The temperature information that temperature sensor is sent is received, water pump is sent control signals to after processing；
Water pump：The control signal that reception processing device is sent is operated.When server temperature raises, for the cold of cooling
But temperature of the water in cooling procedure can also rise, and cooling effectiveness declines, and the flow velocity at this time needing to accelerate cooling water ensures
The reliability of cooling, temperature sensor detect the temperature of delivery port, and the temperature of delivery port is sent into processor, processor according to
The temperature of delivery port carries out PID regulations to water pump, makes the temperature stabilization of delivery port in default scope.
Flowmeter is additionally provided with the water side.Flowmeter can monitor the flow of water side and send flow information
To processor, because flow information can feed back the power of water pump simultaneously, it is easy to processor to be controlled water pump, meanwhile, from stream
The caloric value of server can also be embodied in amount data.
Air-conditioning system is also associated with the fan, the air-conditioning system is connected with processor.In the mistake of cooling water circulation
Cheng Zhong, the round-robin method of full blast is temperature one stable value of holding that cold row exports water, still, when cooling water circulation
When too fast, the effect for depending merely on cold row and fan is not enough to complete the cooling to cooling water, and iterative cycles are to causing the temperature of cooling water
More and more higher is spent, radiating efficiency is more and more lower, in order to avoid such case, when the cooling water flow velocity that flowmeter detects exceedes in advance
If value when, processor opens air-conditioning system, ensures the heat exchange efficiency of cold row.
The processor is also connected with host computer.All information received can be sent to host computer by processor, on
Position machine has staff to be monitored parameter.
The system includes at least one set of server unit, the quantity of the water-cooling joint, water pump and server unit
Match somebody with somebody.One processor can control the radiating of multigroup server, modularized design, be easy to extend.
The utility model compared with prior art, has the following advantages and advantages：
1st, a kind of deep learning system based on rack-mount server of the utility model, water-cooled cooling, cooling effectiveness are high；
2nd, a kind of deep learning system based on rack-mount server of the utility model, intelligent control, the stability of a system is high,
Robustness is good.
Brief description of the drawings
Accompanying drawing described herein is used for providing further understanding the utility model embodiment, forms the one of the application
Part, the restriction to the utility model embodiment is not formed.In the accompanying drawings：
Fig. 1 is the utility model structure diagram.
Mark and corresponding parts title in accompanying drawing：
1- server units, 2- water-cooling joints, 201- water intake ends, 202- water sides, 3- water pumps, the cold rows of 4-, 5- fans, 6-
Processor, 7- temperature sensors, 8- flowmeters, 9- air-conditioning systems, 10- host computers.
Embodiment
For the purpose of this utility model, technical scheme and advantage is more clearly understood, with reference to embodiment and accompanying drawing,
The utility model is described in further detail, and exemplary embodiment of the present utility model and its explanation are only used for explaining this
Utility model, it is not intended as to restriction of the present utility model.
Embodiment
As shown in figure 1, a kind of deep learning system based on rack-mount server of the utility model, including 3 groups of Intel
Xeon E5 server units 1, in addition to the water-cooling joints 2 of EK EVO 10 matched with server unit 1, the water-cooling joint 2
Water intake end 201 by water pump 3 connect the cold row 4 of 480mm fine copper one end, the water side 202 of the water-cooling joint 2 connects cold row
4 other end, 4 pieces of 120mm fan 5 is additionally provided with the cold row 4.Rack-mount server is carried out by way of water cooling
Radiating, in the case where not being extended to space, the efficiency far of water-cooling is higher than wind-cooling heat dissipating.The water-cooling joint 2
Water side 202 on be additionally provided with HX15 temperature sensors 7, be also associated with NXP-S12X processors on the temperature sensor 7
6, processor 6 is also connected with water pump 3；When server temperature raises, temperature of the cooling water in cooling procedure for cooling
Also can rise, cooling effectiveness decline, at this time need accelerate cooling water flow velocity come ensure cooling reliability, temperature sensor
The temperature of 7 detection delivery ports, is sent to processor 6, processor 6 is according to the temperature of delivery port to water pump 3 by the temperature of delivery port
PID regulations are carried out, make the temperature stabilization of delivery port in default scope.FMG70 flows are additionally provided with the water side 202
Meter 8.Flowmeter 8 can monitor the flow of water side 202 and flow information is sent into processor 6, due to flow information simultaneously
The power of water pump 3 can be fed back, is easy to processor 6 to be controlled water pump 3, meanwhile, service can also be embodied from data on flows
The caloric value of device.3kw air-conditioning system 9 is also associated with the fan 5, the air-conditioning system 9 is connected with processor 6.Cold
But during water circulation, the round-robin method of full blast is temperature one stable value of holding that cold row 4 exports water, still,
When cooling water circulation it is too fast when, the effect for depending merely on cold row 4 and fan 5 is not enough to complete the cooling to cooling water, iterative cycles
Temperature more and more higher to causing cooling water, radiating efficiency is more and more lower, in order to avoid such case, when flowmeter 8 detects
Cooling water flow velocity when exceeding default value, processor 6 opens air-conditioning system 9, ensures the heat exchange efficiency of cold row.The processor
6 are also connected by bus with host computer 10.All information received can be sent to host computer 10, host computer by processor 6
10 have staff to be monitored parameter.
Temperature sensor 7：The temperature of the cooling water of water side 202 is detected, temperature information is sent to processor 6；
Processor 6：The temperature information that temperature sensor 7 is sent is received, water pump 3 is sent control signals to after processing；
Water pump 3：The control signal that reception processing device 6 is sent is operated.
Above-described embodiment, the purpose of this utility model, technical scheme and beneficial effect are entered
One step describes in detail, should be understood that and the foregoing is only specific embodiment of the present utility model, is not used to limit
Determine the scope of protection of the utility model, it is all within the spirit and principles of the utility model, any modification for being made, equally replace
Change, improve, should be included within the scope of protection of the utility model.
Claims (6)
1. a kind of deep learning system based on rack-mount server, including server unit (1), it is characterised in that also include
The water-cooling joint (2) matched with server unit (1), the water intake end (201) of the water-cooling joint (2) are connected by water pump (3)
One end of cold row (4), the water side (202) of the water-cooling joint (2) connect the other end of cold row (4), gone back on the cold row (4)
It is provided with fan (5).
A kind of 2. deep learning system based on rack-mount server according to claim 1, it is characterised in that the water
Temperature sensor (7) is additionally provided with the water side (202) of cold junction (2), is also associated with handling on the temperature sensor (7)
Device (6), processor (6) are also connected with water pump (3)；
Temperature sensor (7)：The temperature of water side (202) cooling water is detected, temperature information is sent to processor (6)；
Processor (6)：The temperature information that temperature sensor (7) is sent is received, water pump (3) is sent control signals to after processing；
Water pump (3)：The control signal that reception processing device (6) is sent is operated.
A kind of 3. deep learning system based on rack-mount server according to claim 2, it is characterised in that it is described go out
Flowmeter (8) is additionally provided with water end (W.E.) (202).
A kind of 4. deep learning system based on rack-mount server according to claim 3, it is characterised in that the wind
Air-conditioning system (9) is also associated with fan (5), the air-conditioning system (9) is connected with processor (6).
A kind of 5. deep learning system based on rack-mount server according to claim 2, it is characterised in that the place
Reason device (6) is also connected with host computer (10).
A kind of 6. deep learning system based on rack-mount server according to claim 1, it is characterised in that the system
System includes at least one set of server unit (1), the water-cooling joint (2), the quantity Matching of water pump (3) and server unit (1).
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201720056166.4U CN206805432U (en) | 2017-01-18 | 2017-01-18 | A kind of deep learning system based on rack-mount server |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201720056166.4U CN206805432U (en) | 2017-01-18 | 2017-01-18 | A kind of deep learning system based on rack-mount server |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN206805432U true CN206805432U (en) | 2017-12-26 |
Family
ID=60729420
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201720056166.4U Expired - Fee Related CN206805432U (en) | 2017-01-18 | 2017-01-18 | A kind of deep learning system based on rack-mount server |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN206805432U (en) |
-
2017
- 2017-01-18 CN CN201720056166.4U patent/CN206805432U/en not_active Expired - Fee Related
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US7421623B2 (en) | Systems, methods, and media for controlling temperature in a computer system | |
| CN101853059B (en) | Cloud computing server system for heat dissipation, energy conservation and safe data storage | |
| CN105975029A (en) | Distributed temperature control case radiating system | |
| CN205788082U (en) | The enclosure heat radiation system that a kind of distributed temperature controls | |
| US10390463B2 (en) | Backflow prevention for computing devices | |
| US20120118534A1 (en) | Multimodal cooling apparatus for an electronic system | |
| US9655286B2 (en) | Facilitating fluid level sensing | |
| CN102052338A (en) | Server framework and fan system thereof | |
| CN104582416A (en) | Liquid cooling system and control method thereof | |
| CN102253700A (en) | Fan control system | |
| JP2014183062A (en) | Electronic apparatus cooling system and electronic apparatus cooling method | |
| US7006949B2 (en) | Method and system for collecting temperature data | |
| CN201364544Y (en) | High-end cabinet server radiating structure device | |
| US8315054B2 (en) | Rack server center | |
| CN106919490A (en) | Server failure detection method and device | |
| US8325482B2 (en) | Cooling apparatus for server rack | |
| CN206805432U (en) | A kind of deep learning system based on rack-mount server | |
| CN211267493U (en) | All-in-one machine and efficient heat dissipation cabinet thereof | |
| CN106919533B (en) | 4U high-density storage type server | |
| US20120257348A1 (en) | Data center and heat dissipating system thereof | |
| Meyer et al. | iDataCool: HPC with hot-water cooling and energy reuse | |
| TW201321610A (en) | Fan control system and method thereof | |
| CN106852074A (en) | A kind of intelligent learning system based on multiserver | |
| US20140177097A1 (en) | Hard disk drive module having temperature detecting function | |
| CN202815702U (en) | Heat radiation system for computer server |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20171226 Termination date: 20200118 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |