CN108762818A - A kind of optimization design server and maintaining method - Google Patents
A kind of optimization design server and maintaining method Download PDFInfo
- Publication number
- CN108762818A CN108762818A CN201810538595.4A CN201810538595A CN108762818A CN 108762818 A CN108762818 A CN 108762818A CN 201810538595 A CN201810538595 A CN 201810538595A CN 108762818 A CN108762818 A CN 108762818A
- Authority
- CN
- China
- Prior art keywords
- server
- gpu
- optimization design
- traditional
- layer
- 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.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/44—Arrangements for executing specific programs
- G06F9/4401—Bootstrapping
- G06F9/4411—Configuring for operating with peripheral devices; Loading of device drivers
- G06F9/4413—Plug-and-play [PnP]
-
- H—ELECTRICITY
- H05—ELECTRIC TECHNIQUES NOT OTHERWISE PROVIDED FOR
- H05K—PRINTED CIRCUITS; CASINGS OR CONSTRUCTIONAL DETAILS OF ELECTRIC APPARATUS; MANUFACTURE OF ASSEMBLAGES OF ELECTRICAL COMPONENTS
- H05K5/00—Casings, cabinets or drawers for electric apparatus
- H05K5/02—Details
- H05K5/0217—Mechanical details of casings
Landscapes
- Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Theoretical Computer Science (AREA)
- Computer Security & Cryptography (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Microelectronics & Electronic Packaging (AREA)
- Power Sources (AREA)
Abstract
The invention discloses a kind of optimization design server and maintaining methods, the server of optimized design, traditional server system and GPU server systems are layered according to function module and are arranged, and layering quantity is multilayer, can freely stretch stretching in the case of realization complete machine uninterrupted operation, and the side of each layer is equipped with the tank chain for pull traditional server system or GPU server systems in addition to top layer, it ensure that the arrangement continuity between each layer, it avoids in pull layer, the loss of parts, and it is equipped with fixing buckle in the other side of each layer, strengthen the structural integrity of server after optimization design.
Description
Technical field
The present invention relates to artificial intelligence servers fields, more particularly, to a kind of optimization design server and maintaining method.
Background technology
Artificial intelligence plays increasingly important role in the online service field of our times, Google,
These intra-companies of Facebook, Microsoft and Baidu, GPU play huge effect in " deep learning " field.GPU is one
Large-scale SIMD set, possesses a large amount of vector operation system, can be with parallel processing when carrying out the stream calculation of repeat function
A large amount of trifling information, it is only necessary to set performance data constantly inward into can.The powerful of CPU is in face of control intensity
It is very outstanding when operation, but for data-intensive operation, for example each pixel should show any color in screen,
Comparatively the operating mechanism pressure that may be restricted to CPU can be very big.
With flourishing for artificial intelligence, GPU is largely widely used, and for large-scale data center, is had begun
The server of a large amount of deployment pond classes, this kind of traditional server does not support the hot O&M of GPU in structure, once server needs
It overhauls or GPU upgrades, need entirety that server is carried out power-off and carry out relevant operation, the no image of Buddha replaces hard disk, wind
Fan, the such operation readiness of power supply, this can cause entire service stopping to operate, and bring great challenge to whole system, therefore anxious
Need to design and develop it is a kind of can the server of hot O&M and easily upgrading GPU meet the market demand, meet entire service.
Invention content
To solve the above problems, the present invention provides a kind of optimization design server, the server GPU branch after optimization design
Hold hot O&M, it can be achieved that single part maintenance, the maintaining method is convenient and efficient.
Based on this, the technical scheme is that：
A kind of optimization design server, including traditional server system, GPU server systems, tank chain, electric wire and fixation
It is multilayer that button, the traditional server system and GPU server systems are layered setting and layering quantity according to function module, is removed
The first side of each layer is equipped with the tank chain for pull traditional server system or GPU server systems, institute outside top layer
It states electric wire to be interspersed in tank chain, fixing buckle is equipped in the second side of each layer.
Further, the traditional server system include CPU processor touch block, memory module, network module, power supply and
Fan.
Further, multiple GPU processor modules are equipped in the GPU server systems.
Further, the GPU processor dies number of blocks is 6.
Further, the traditional server system and GPU server systems are independently disposed in different layers.
Further, the electric wire is power supply electric wire and data service cabling.
In addition, the present invention also provides a kind of maintaining method of optimization design server, the clothes of above-mentioned optimization design are utilized
Business device, when the server carries out upgrading or O&M is replaced,
A. enter traditional server system and disable falls corresponding GPU server systems；
B. releasing fixing buckle will need to change or the GPU server systems of O&M pull out, and replacement is pushed back after successfully replacing
GPU server systems afterwards；
C., fixing buckle is installed, traditional server system is again introduced into and updates driving again and identify corresponding GPU services
Device system, then the server enter normal traffic pattern.
Implement the embodiment of the present invention, has the advantages that：
1, the server of the optimized design of the present invention, by traditional server system and GPU server systems according to function mould
Block is layered setting, and it is multilayer to be layered quantity, and can freely stretch stretching in the case of realization complete machine uninterrupted operation, and remove top layer
The side of outer each layer is equipped with the tank chain for pull traditional server system or GPU server systems, ensure that each
Arrangement continuity between a layer is avoided in pull layer, the loss of parts, and is equipped with and is fixed in the other side of each layer
Button, strengthens the structural integrity of server after optimization design.
2, multiple GPU processor modules are equipped in GPU server systems of the invention, are preferable over 6 GPU processors, into
When row artificial intelligence, deep learning, neural network are deduced, working efficiency can be greatly promoted, efficiency can more biographies shoulder to shoulder
System server.
3, the traditional server system and GPU server systems are independently disposed in different layers, are independently arranged
Facilitate later maintenance；In addition electric wire is power supply electric wire and data service cabling, and by being interspersed in tank chain and traditional server
System or the connection of GPU server systems, realize access of the server CPU to GPU resource.
4, the method clear and rational safeguarded using the server of the present invention is carrying out O&M or iteration to the server
Upgrading, realizes the uninterrupted operation of the server, avoids loss of data.
Description of the drawings
Fig. 1 is the overall structure diagram of optimization design server described in the present embodiment.
Fig. 2 is the flow chart for carrying out maintaining method described in the present embodiment using above-mentioned optimization design server.
Reference sign：
Wherein, 1, traditional server system, 2, GPU server systems, 3, tank chain, 4, fixing buckle.
Specific implementation mode
Below in conjunction with the drawings and examples in the present invention, technical scheme of the present invention is clearly and completely retouched
It states, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.Based on the present invention
In embodiment, every other implementation obtained by those of ordinary skill in the art without making creative efforts
Example, belongs to protection scope of the present invention.
In conjunction with shown in Fig. 1, a kind of optimization design server described in the embodiment of the present invention, including traditional server system 1,
GPU server systems 2, tank chain 3, the electric wire and fixing buckle 4, traditional server system 1 and GPU server systems 2 according to
Function module layering is arranged and layering quantity is multilayer, and the first side of each layer is equipped with for pull tradition in addition to top layer
The tank chain 3 of server system 1 or GPU server systems 2, the electric wire are interspersed in tank chain 3, the second of each layer
Side is equipped with fixing buckle 4.The server of the optimized design of the embodiment of the present invention, by traditional server system 1 and GPU servers system
System 2 is layered according to function module to be arranged, and it is multilayer to be layered quantity, and specific each layer is that traditional server system 1 or GPU are serviced
Device system 2 is not specific, and can freely stretch stretching, and each layer in addition to top layer in the case of realization complete machine uninterrupted operation
Side is equipped with the tank chain 3 for pull traditional server system 1 or GPU server systems 2, ensure that between each layer
Continuity is arranged, is avoided in pull layer, the loss of parts, and fixing buckle 4 is equipped in the other side of each layer, is strengthened
The structural integrity of server after optimization design.
Wherein, traditional server system 1 touches block, memory module, network module, power supply and blower including CPU processor,
It is equipped with multiple GPU processor modules in GPU server systems 2, is preferable over 6 GPU processors, carries out artificial intelligence, depth
It practises, when neural network is deduced, can greatly promote working efficiency, efficiency can more traditional servers shoulder to shoulder.Traditional services
Device system 1 and GPU server systems 2 are independently disposed in different layers, are independently arranged and are facilitated later maintenance；In addition the electricity
Line is power supply electric wire and data service cabling, and by being interspersed in tank chain 3 and traditional server system 1 or GPU servers system
2 connection of system, realizes access of the server CPU to GPU resource.
In addition, using tank chain 3 in the side of each layer in addition to top layer, tank chain 3, which plays, to be suitable for use in back and forth
The occasion of movement can play traction and protective effect to the built-in electric wire；Tank chain 3 often save by left and right carrier bar and up and down
Cover board forms, and often section can be opened and be rotated freely between chain link, be easy for installation and maintenance, be convenient to mount and dismount, need not thread, and opens
Cable can be conveniently installed inside after cover board, greatly improve the convenient degree of cabling；Tank chain 3 has higher pressure
Power and tension load, good toughness, high resiliency and wearability are fire-retardant, and performance is stablized when high/low temperature, can use in outdoor,
The speed of service is stablized.
In conjunction with shown in Fig. 2, the present invention also provides a kind of maintaining methods of optimization design server, are set using above-mentioned optimization
The server of meter, when the server carries out upgrading or O&M is replaced,
A. enter traditional server system 1 and disable falls corresponding GPU server systems 2；
B. releasing fixing buckle 4 will need to change or the GPU server systems 2 of O&M pull out, and replacement is pushed back after successfully replacing
GPU server systems 2 later；
C., fixing buckle is installed, traditional server system 1 is again introduced into and updates driving again and identify corresponding GPU services
Device system 2, then the server enter normal traffic pattern.
The method clear and rational safeguarded using the server of the present invention is carrying out O&M or iteration liter to the server
Grade, realizes the uninterrupted operation of the server, avoids loss of data.
The above is only a preferred embodiment of the present invention, it should be pointed out that for those skilled in the art
For, without departing from the technical principles of the invention, several improvement and replacement can also be made, these improve and replace
It should be regarded as protection scope of the present invention.
Claims (7)
1. a kind of optimization design server, which is characterized in that including traditional server system, GPU server systems, tank chain,
Electric wire and fixing buckle, the traditional server system and GPU server systems are layered setting and layering quantity according to function module
For multilayer, the first side of each layer is equipped with for pull traditional server system or GPU server systems in addition to top layer
Tank chain, the electric wire are interspersed in tank chain, and fixing buckle is equipped in the second side of each layer.
2. optimization design server as described in claim 1, which is characterized in that the traditional server system includes CPU processing
Device touches block, memory module, network module, power supply and blower.
3. optimization design server as described in claim 1, which is characterized in that be equipped with multiple GPU in the GPU server systems
Processor module.
4. optimization design server as claimed in claim 3, which is characterized in that the GPU processor dies number of blocks is 6.
5. optimization design server as described in claim 1, which is characterized in that the traditional server system and GPU servers
System is independently disposed in different layers.
6. optimization design server as described in claim 1, which is characterized in that the electric wire is that power supply electric wire and data service are walked
Line.
7. a kind of maintaining method of optimization design server utilizes the service of any one of claim 1 to 6 optimization design
Device, which is characterized in that when the server carries out upgrading or O&M is replaced,
A. enter traditional server system and disable falls corresponding GPU server systems；
B. releasing fixing buckle will need to upgrade or the GPU server systems of O&M pull out, and be pushed back after successfully replacing after replacing
GPU server systems；
C., fixing buckle is installed, traditional server system is again introduced into and updates driving again and identify corresponding GPU servers system
System, then the server enters normal traffic pattern.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810538595.4A CN108762818A (en) | 2018-05-30 | 2018-05-30 | A kind of optimization design server and maintaining method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810538595.4A CN108762818A (en) | 2018-05-30 | 2018-05-30 | A kind of optimization design server and maintaining method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108762818A true CN108762818A (en) | 2018-11-06 |
Family
ID=64004001
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810538595.4A Pending CN108762818A (en) | 2018-05-30 | 2018-05-30 | A kind of optimization design server and maintaining method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108762818A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112214295A (en) * | 2020-09-23 | 2021-01-12 | 桂林理工大学 | Low-energy-consumption job scheduling method for multi-CPU/GPU heterogeneous server cluster |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101896940A (en) * | 2007-10-10 | 2010-11-24 | 苹果公司 | Framework for dynamic configuration of hardware resources |
| CN103389777A (en) * | 2013-06-26 | 2013-11-13 | 汉柏科技有限公司 | Storage server |
| CN203338221U (en) * | 2013-05-15 | 2013-12-11 | 汉柏科技有限公司 | Server case |
| CN104125165A (en) * | 2014-08-18 | 2014-10-29 | 浪潮电子信息产业股份有限公司 | Job scheduling system and method based on heterogeneous cluster |
| CN106249818A (en) * | 2016-08-01 | 2016-12-21 | 浪潮电子信息产业股份有限公司 | A kind of server node and system |
-
2018
- 2018-05-30 CN CN201810538595.4A patent/CN108762818A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101896940A (en) * | 2007-10-10 | 2010-11-24 | 苹果公司 | Framework for dynamic configuration of hardware resources |
| CN203338221U (en) * | 2013-05-15 | 2013-12-11 | 汉柏科技有限公司 | Server case |
| CN103389777A (en) * | 2013-06-26 | 2013-11-13 | 汉柏科技有限公司 | Storage server |
| CN104125165A (en) * | 2014-08-18 | 2014-10-29 | 浪潮电子信息产业股份有限公司 | Job scheduling system and method based on heterogeneous cluster |
| CN106249818A (en) * | 2016-08-01 | 2016-12-21 | 浪潮电子信息产业股份有限公司 | A kind of server node and system |
Non-Patent Citations (1)
| Title |
|---|
| 刘纪红，潘学俊，梅栴等编著: "《物联网技术及应用》", 31 December 2011, 国防工业出版社 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112214295A (en) * | 2020-09-23 | 2021-01-12 | 桂林理工大学 | Low-energy-consumption job scheduling method for multi-CPU/GPU heterogeneous server cluster |
| CN112214295B (en) * | 2020-09-23 | 2024-02-06 | 桂林理工大学 | Low-energy-consumption job scheduling method for multi-CPU/GPU heterogeneous server cluster |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101216777B (en) | Rapid deployment system under multi-dummy machine environment | |
| CN109313582A (en) | Technology for dynamic remote resource allocation | |
| US7370156B1 (en) | Unity parallel processing system and method | |
| KR20120092579A (en) | High density multi node computer with integrated shared resources | |
| CN106547592A (en) | A kind of method for designing for realizing Novel cloud service device power supply software online updating | |
| CN111327692A (en) | Model training method and device and cluster system | |
| CN108023958A (en) | A kind of resource scheduling system based on cloud platform resource monitoring | |
| CN109791493A (en) | System and method for the load balance in the decoding of out-of-order clustering | |
| CN108762818A (en) | A kind of optimization design server and maintaining method | |
| CN105007311A (en) | System and method for resource management based on cloud platform and cloud computing | |
| CN106648775A (en) | Double-open switching method and system for application program and terminal equipment | |
| CN203786606U (en) | Cabinet type server device | |
| CN107918560A (en) | A kind of server apparatus management method and device | |
| CN109032422A (en) | Intelligent handwriting board and implementation method and device thereof | |
| CN205992328U (en) | A kind of cloud Classroom System based on cloud service | |
| CN204117069U (en) | A kind of server backplanes | |
| CN109688068A (en) | Network load balancing method and device based on big data analysis | |
| CN108984471A (en) | A kind of ether mill mine machine system and its dig mine method | |
| CN109739560A (en) | A kind of GPU card cluster configuration control system and method | |
| CN102096847A (en) | Information system based on multistage centralized and hierarchical management mode | |
| CN109495484B (en) | High-concurrency communication method and system for Proactor mode in intelligent lighting system | |
| CN106294062A (en) | A kind of method by computing function in server with management function separate design | |
| CN207166346U (en) | A kind of power module mounting structure | |
| Xiong et al. | A novel think tanks evaluation system based on micro service | |
| CN109189489A (en) | A method of it solving distributed memory system and restarts rear cluster network interface card confusion |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181106 |