CN107239346A - A kind of whole machine cabinet computing resource tank node and computing resource pond framework - Google Patents
A kind of whole machine cabinet computing resource tank node and computing resource pond framework Download PDFInfo
- Publication number
- CN107239346A CN107239346A CN201710433600.0A CN201710433600A CN107239346A CN 107239346 A CN107239346 A CN 107239346A CN 201710433600 A CN201710433600 A CN 201710433600A CN 107239346 A CN107239346 A CN 107239346A
- Authority
- CN
- China
- Prior art keywords
- gpu
- node
- whole machine
- computing resource
- machine cabinet
- 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
- G06F9/46—Multiprogramming arrangements
- G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
- G06F9/5005—Allocation of resources, e.g. of the central processing unit [CPU] to service a request
- G06F9/5027—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resource being a machine, e.g. CPUs, Servers, Terminals
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2209/00—Indexing scheme relating to G06F9/00
- G06F2209/50—Indexing scheme relating to G06F9/50
- G06F2209/5011—Pool
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Exchanges In Wide-Area Networks (AREA)
Abstract
The invention discloses a kind of whole machine cabinet computing resource tank node and computing resource pond framework, the whole machine cabinet server for being configured with management module, calculate node is applied to the form of 1U nodes, its structure includes power panel, GPU node modules and GPU, the GPU node modules are connected to above-mentioned management module by power panel, realize to GPU node modules condition monitoring and the management function of computing resource;Data exchange chip is configured with GPU node modules, the data exchange chip can connect the calculate node, GPU and realize that data are calculated between GPU and calculate node to be exchanged.A kind of whole machine cabinet computing resource tank node and computing resource pond framework of the present invention is compared with prior art, pass through upper layer software (applications) pond administrative skill, realize that the dynamic pondization of resource and task are distributed automatically, reach that the maximization of node resource is used, improve resource pool flexibility, utilization rate, system energy consumption is reduced, it is practical.
Description
Technical field
The present invention relates to field of computer technology, specifically a kind of whole machine cabinet computing resource tank node and computing resource
Pond framework.
Background technology
With the fast development of internet economy, mass data just impacts whole data with unprecedented growth trend
Center industry, higher requirement is proposed to IT infrastructure.Server is as one of core component of data center, in order to suitable
The demand for answering following extensive business to increase, it is also desirable to which its framework is optimized and reconstructed.
In the Resources re engineering framework of server, computing resource reconstruct is one of important application.Simultaneously modularization and
High density is the important trend of server development, shows as generic server and gradually develops to whole machine cabinet server.
The pondization design of current computing resource is not applied to whole machine cabinet server field, and integration density is low, high energy consumption, nothing
Method is managed concentratedly, and resource allocation mode ossifys, and utilization of resources rate is low, installs and maintenance workload is big.
Based on this, the present invention provides a kind of whole machine cabinet computing resource tank node and computing resource pond framework.Solve complete machine
Cabinet computing resource pond architecture design technology, forms 1U node modules by computing resource pond and is applied to whole machine cabinet server, and
Realize the functions such as cascade extension, dynamic pond, centralized management.
The content of the invention
The technical assignment of the present invention is that there is provided a kind of whole machine cabinet computing resource tank node and calculating for above weak point
Resource pool framework.
A kind of whole machine cabinet computing resource tank node, is applied to the form of 1U nodes and is configured with management module, calculate node
Whole machine cabinet server, its structure includes power panel, GPU node modules and GPU, and the GPU node modules are connected by power panel
Above-mentioned management module is connected to, is realized to GPU node modules condition monitoring and the management function of computing resource;In GPU node modules
In be configured with data exchange chip, the data exchange chip can connect the calculate node, GPU and realize GPU and calculate node
Between calculate data exchange.
Powered between power panel and the GPU node module using copper bar, supply voltage is 12V.
The data exchange chip configures 2 data upstream Interfaces and 4 data downstream interfaces, 4 data downstream interfaces
4 GPU are respectively connected to, 1 data upstream Interface can access calculate node, and the data uplink interface, data downstream interface are
PCIE interfaces.
BMC chip, MCPU chips and the PCIE Switch chips of order interconnection are also configured with the GPU node modules,
The PCIE Switch chips connect above-mentioned data exchange chip and are also associated with expansible external management interface, this pair of outer tube
Reason interface is PCIE interfaces.
The computing resource tank node can be used for cascading, i.e., interconnect at least two GPU node modules, specific cascade structure
For:The upstream Interface of a GPU node modules is accessed into calculate node first, another upstream Interface of the GPU node modules then connects
Enter a upstream Interface of another GPU node modules;External management interface between two GPU node modules is interconnected, and realizes
PCIE manages the intercommunication of signal;It is then real using above-mentioned connected mode between another GPU node modules and other GPU node modules
Now cascade.
In the GPU node modules being connected with calculate node, the management to GPU node modules, MCPU are realized by MCPU
External management interface and data exchange chip are connected to by 1 PCIE Switch chip, ascending tube is realized by BMC chip
The dynamic select that passage is 1 and 2 is managed, that is, which data uplink interface selected, when calculate node module is to be cascaded module, pipe
Link switching is managed to passage 1, keeps 1 MCPU to carry out the management of 2 or N number of GPU node modules, N here is to be cascaded mould
The quantity of block, so as to realize the cascade of GPU node modules.
A kind of whole machine cabinet computing resource pond framework, including calculate node, some GPU node modules, a whole machine cabinet pipe
Module and whole machine cabinet power bus BUSBAR are managed, calculate node and GPU node modules are connected to by respective power panel respectively
Whole machine cabinet power bus BUSBAR power takings, realize the centrally connected power supply in computing resource pond;Whole machine cabinet management module is used to realize to whole
The centralized management in rack computing resource pond, calculate node is used for the main equipment end as computing resource pond, is connected respectively by cable
It is connected to each GPU node modules and transmits PCIE data-signals.
BMC chip in the calculate node, GPU nodes is logical by respective power panel and whole machine cabinet management module respectively
Letter, so as to realize the centralized management in computing resource pond;The whole machine cabinet management module is used to collect calculate node and GPU node modules
Resource information, resource utilization, and report the upper application software in the whole machine cabinet management module.
The whole machine cabinet management module communicated with monitoring chip BMC obtain resource information include cpu busy percentage, GPU profit
With rate, the network bandwidth, and resource utilization in resource pool is reported into upper application software in time.
All GPU resource Unified codings of acquisition, management are formed GPU resource pond by the system upper application software, and
According to specific related resource utilization rate, the business saturation degree of each GPU in GPU resource pond, effective adjustresources pool service are calculated
Using realizing the dynamic pond of resource, while new processor active task can be distributed automatically, realize that the maximization of node resource is used.
Compared to the prior art a kind of whole machine cabinet computing resource tank node and computing resource pond framework of the present invention, have
Following beneficial effect:
1), computing resource tank node module whole machine cabinet server is applied to the form of 1U nodes, improve deployment density.
2), whole machine cabinet computing resource pond can realize centrally connected power supply, centralized management, improve efficiency, reduce system energy consumption.
3), GPU node modules can realize that data are cascaded, and the dynamic of link management can be realized by BMC chip, reach meter
Resource pool extension purpose is calculated, calculate node resource requirement is reduced, reduces cost.
4), based on the design of computing resource pond node module, with reference to calculate node, build the resource pool of whole machine cabinet form
Framework, realizes the centrally connected power supply of whole machine cabinet form, centralized management, dynamic pond, raising delivery efficiency, O&M efficiency.
5), system upper application software by all GPU resource Unified codings in the mechanism, management, form GPU resource pond,
And according to specific related resource utilization rate, calculate each GPU business saturation degree, effective adjustresources pool service in GPU resource pond
Using, the dynamic pond of resource is realized, while new processor active task can be distributed automatically, realizes that the maximization of node resource is used, so that
Resource pool flexibility, utilization rate are improved, system energy consumption is reduced, it is practical, it is applied widely, with good popularization and application
Value.
Brief description of the drawings
Accompanying drawing 1 is whole machine cabinet computing resource tank node schematic diagram.
Accompanying drawing 2 is GPU node module cascade schematic diagrames.
Accompanying drawing 3 is computing resource pond link management cascade schematic diagram.
Accompanying drawing 4 is whole machine cabinet computing resource pond configuration diagram.
Embodiment
Below in conjunction with the accompanying drawings and specific embodiment the invention will be further described.
A kind of whole machine cabinet computing resource tank node, by the computing resource tank node module, with the form application of 1U nodes
To whole machine cabinet server, the centralized management in computing resource pond can be achieved, integration density, reduction energy consumption is improved.Save in computing resource pond
4 GPU of point can be directly realized by mutual calculating data exchange by such as PEX9797 data exchange chip, the node module
Data-interface can be cascaded to other 1 computing resource tank node, realize the cascade of computing resource pond data exchange unit.In number
During according to cascade, the switching at runtime of link management is realized by BMC chip, keeps 1 MCPU to carry out 2 computing resource data exchanges
The management of unit, realizes the cascade of data exchange administrative unit.
Whole machine cabinet management system collects the resource information of calculate node and GPU nodes, resource utilization and reports upper strata
Application software.All GPU resource Unified codings, management in the mechanism are formed GPU resource pond by system upper application software, and
According to specific related resource utilization rate, each GPU business saturation degree in GPU resource pond is calculated, effective adjustresources pool service should
With, the dynamic pond of resource is realized, and new processor active task is distributed automatically, realize that the maximization of node resource is used.
As shown in Figure 1, concrete structure of the invention includes power panel, GPU node modules and GPU, the GPU nodes mould
Block is connected to above-mentioned management module by power panel, realizes to GPU node modules condition monitoring and the management function of computing resource;
Data exchange chip is configured with GPU node modules, the data exchange chip can connect the calculate node, GPU and realize
Exchanging for data is calculated between GPU and calculate node.
Powered between power panel and the GPU node module using copper bar, the copper bar is mutual with node module by power panel
Connection realizes 12V system power supplies, and hot plug, excessively stream, overvoltage circuit design are carried out to power panel, node module system is improved reliable
Property.
The data exchange chip configures 2 data upstream Interfaces and 4 data downstream interfaces, 4 data downstream interfaces
4 GPU are respectively connected to, 1 data upstream Interface can access calculate node, and the data uplink interface, data downstream interface are
PCIE interfaces.
BMC chip, MCPU chips and the PCIE Switch chips of order interconnection are also configured with the GPU node modules,
The PCIE Switch chips connect above-mentioned data exchange chip and are also associated with expansible external management interface, this pair of outer tube
Reason interface is PCIE interfaces.
The computing resource tank node can be used for cascading, i.e., interconnect at least two GPU node modules, specific cascade structure
For:The upstream Interface of a GPU node modules is accessed into calculate node first, another upstream Interface of the GPU node modules then connects
Enter a upstream Interface of another GPU node modules;External management interface between two GPU node modules is interconnected, and realizes
PCIE manages the intercommunication of signal;It is then real using above-mentioned connected mode between another GPU node modules and other GPU node modules
Now cascade.
In the GPU node modules being connected with calculate node, the management to GPU node modules, MCPU are realized by MCPU
External management interface and data exchange chip are connected to by 1 PCIE Switch chip, realized by BMC chip up
Management passage is 1 and 2 dynamic select, that is, selects which data uplink interface, when calculate node module is to be cascaded module,
Link management is switched to passage 1, keeps 1 MCPU to carry out the management of 2 or N number of GPU node modules, and N here is to be cascaded
The quantity of module, so as to realize the cascade of GPU node modules.
Data exchange chip is by taking PEX9797 chips as an example, as shown in Fig. 2 being computing resource tank node module-cascade framework
Schematic diagram.4 GPU of computing resource tank node can be directly realized by mutual calculating data exchange, node by PEX9797 chips
1 data upstream Interface of module is connected to calculate node, and another 1 data-interface is connected to other 1 computing resource tank node,
Realize the cascade of computing resource pond data exchange unit.
As shown in figure 3, being computing resource pond link management cascade schematic diagram.Computing resource tank node module passes through
MCPU realizes the management of computing resource data exchange unit, and MCPU PCIEx1 management signals pass through 1 PCIE Switch chip
Be connected to external management interface and data exchange chip PEX9797, by BMC chip realize up management passage for 1 and 2 it is dynamic
State is selected.When calculate node module for when being cascaded module, link management is switched to passage 1,1 MCPU is kept to carry out 2 meters
The management of resource data crosspoint is calculated, the cascade of data exchange administrative unit is realized.
A kind of whole machine cabinet computing resource pond framework, as shown in figure 4, its structure includes a calculate node, some GPU sections
Point module, whole machine cabinet management module and whole machine cabinet power bus BUSBAR, calculate node and GPU node modules are respectively by respective
Power panel be connected to whole machine cabinet power bus BUSBAR power takings, realize the centrally connected power supply in computing resource pond;Whole machine cabinet manages mould
Block is used to realize the centralized management to whole machine cabinet computing resource pond, and calculate node passes through as the Host ends in computing resource pond
PCIERedriver chips strengthen PCIE signal driving force, and PCIE data-signals are connected respectively to each GPU nodes by cable
Module, forms GPU resource pond, realizes whole machine cabinet computing resource pond.
BMC chip in the calculate node, GPU nodes is logical by respective power panel and whole machine cabinet management module respectively
Letter, so as to realize the centralized management in computing resource pond;The whole machine cabinet management module is used to collect calculate node and GPU node modules
Resource information, resource utilization, and report the upper application software in the whole machine cabinet management module.
The whole machine cabinet management module communicated with monitoring chip BMC obtain resource information include cpu busy percentage, GPU profit
With rate, the network bandwidth, and resource utilization in resource pool is reported into upper application software in time.
All GPU resource Unified codings of acquisition, management are formed GPU resource pond by the system upper application software, and
According to specific related resource utilization rate, the business saturation degree of each GPU in GPU resource pond, effective adjustresources pool service are calculated
Using realizing the dynamic pond of resource, while new processor active task can be distributed automatically, realize that the maximization of node resource is used.
In the present invention, computing resource high-speed data crosspoint is built based on data exchange chip, forms computing resource
Tank node module, whole machine cabinet server is applied to the form of 1U nodes, and the centralized management in computing resource pond can be achieved, collection is improved
Into density, reduction energy consumption.
Computing resource tank node module data cascade is built according to Fig. 2, reaches that computing resource pond extends purpose, reduces
Calculate node resource requirement, reduces cost.
Computing resource tank node module management link cascade is built according to Fig. 3, computing resource pond expansion management is realized
Demand, reduces cost.
According to shown in Fig. 4, based on the design of computing resource pond node module, with reference to calculate node, whole machine cabinet form is built
Resource pool framework, realize the centrally connected power supply of whole machine cabinet form, centralized management, dynamic pond, improve delivery efficiency, O&M and imitate
Rate.
By upper layer software (applications) pond administrative skill, realize that the dynamic pondization of resource and task are distributed automatically, reach node resource
Maximization use, improve resource pool flexibility, utilization rate, reduce system energy consumption.
So as to realize the dynamic pond framework of whole machine cabinet computing resource for supporting cascade.
The technical program is also used in server and storage mainboard plant produced test stage, for BIOS, BMC, CPLD
Version checking.
By embodiment above, the those skilled in the art can readily realize the present invention.But should
Work as understanding, the present invention is not limited to above-mentioned embodiment.On the basis of disclosed embodiment, the technical field
Technical staff can be combined different technical characteristics, so as to realize different technical schemes.
It is the known technology of those skilled in the art in addition to the technical characteristic described in specification.
Claims (10)
1. a kind of whole machine cabinet computing resource tank node, it is characterised in that be applied to the form of 1U nodes be configured with management module,
The whole machine cabinet server of calculate node, its structure includes power panel, GPU node modules and GPU, and the GPU node modules pass through
Power panel is connected to above-mentioned management module, realizes to GPU node modules condition monitoring and the management function of computing resource;In GPU
Data exchange chip is configured with node module, the data exchange chip can connect the calculate node, GPU and realize GPU with
The exchange of data is calculated between calculate node.
2. a kind of whole machine cabinet computing resource tank node according to claim 1, it is characterised in that the power panel and GPU
Powered between node module using copper bar, supply voltage is 12V.
3. a kind of whole machine cabinet computing resource tank node according to claim 1, it is characterised in that the data exchange chip
2 data upstream Interfaces and 4 data downstream interfaces are configured, 4 data downstream interfaces are respectively connected in 4 GPU, 1 data
Line interface can access calculate node, and the data uplink interface, data downstream interface are PCIE interfaces.
4. a kind of whole machine cabinet computing resource tank node according to claim 3, it is characterised in that the GPU node modules
In be also configured with order interconnection BMC chip, MCPU chips and PCIE Switch chips, the PCIE Switch chips connection on
State data exchange chip and be also associated with expansible external management interface, the external management interface is PCIE interfaces.
5. a kind of whole machine cabinet computing resource tank node according to claim 4, it is characterised in that the computing resource pond section
Point can be used for cascading, i.e., interconnect at least two GPU node modules, and specific cascade structure is:First by a GPU node modules
Upstream Interface accesses calculate node, and it is one up that another upstream Interfaces of the GPU node modules then accesses another GPU node modules
Interface;External management interface between two GPU node modules is interconnected, and realizes that PCIE manages the intercommunication of signal;Another GPU
Cascade is then realized using above-mentioned connected mode between node module and other GPU node modules.
6. a kind of whole machine cabinet computing resource tank node according to claim 4 or 5, it is characterised in that with calculate node
In the GPU node modules of connection, the management to GPU node modules is realized by MCPU, MCPU chips pass through 1 PCIE
Switch chips are connected to external management interface and data exchange chip, realize that up management passage is 1 and 2 by BMC chip
Dynamic select, that is, which data uplink interface selected, when calculate node module for be cascaded module when, link management is switched to
Passage 1, keeps 1 MCPU to carry out the management of 2 or N number of GPU node modules, N here is the quantity for being cascaded module, so that
Realize the cascade of GPU node modules.
7. a kind of whole machine cabinet computing resource pond framework, it is characterised in that including a calculate node, some GPU node modules,
Whole machine cabinet management module and whole machine cabinet power bus BUSBAR, calculate node and GPU node modules pass through respective power supply respectively
Plate is connected to whole machine cabinet power bus BUSBAR power takings, realizes the centrally connected power supply in computing resource pond;Whole machine cabinet management module is used for
The centralized management to whole machine cabinet computing resource pond is realized, calculate node is used for the main equipment end as computing resource pond, passes through line
Cable is connected respectively to each GPU node modules and transmits PCIE data-signals.
8. a kind of whole machine cabinet computing resource pond framework according to claim 7, it is characterised in that the calculate node,
BMC chip in GPU nodes is communicated by respective power panel with whole machine cabinet management module respectively, so as to realize computing resource pond
Centralized management;The whole machine cabinet management module is used for resource information, the utilization of resources for collecting calculate node and GPU node modules
Rate, and report the upper application software in the whole machine cabinet management module.
9. a kind of whole machine cabinet computing resource pond framework according to claim 8, it is characterised in that the whole machine cabinet management
The module resource information obtained that communicated with monitoring chip BMC includes cpu busy percentage, GPU utilization rates, the network bandwidth, and by resource
Resource utilization reports upper application software in time in pond.
10. a kind of whole machine cabinet computing resource pond framework according to claim 9, it is characterised in that the system upper strata
All GPU resource Unified codings of acquisition, management are formed GPU resource pond by application software, and according to specific related resource profit
With rate, the business saturation degree of each GPU in GPU resource pond is calculated, the dynamic pond of resource is realized in effective adjustresources pool service application
Change, while new processor active task can be distributed automatically, realize that the maximization of node resource is used.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710433600.0A CN107239346A (en) | 2017-06-09 | 2017-06-09 | A kind of whole machine cabinet computing resource tank node and computing resource pond framework |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710433600.0A CN107239346A (en) | 2017-06-09 | 2017-06-09 | A kind of whole machine cabinet computing resource tank node and computing resource pond framework |
Publications (1)
Publication Number | Publication Date |
---|---|
CN107239346A true CN107239346A (en) | 2017-10-10 |
Family
ID=59986082
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710433600.0A Pending CN107239346A (en) | 2017-06-09 | 2017-06-09 | A kind of whole machine cabinet computing resource tank node and computing resource pond framework |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107239346A (en) |
Cited By (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107748726A (en) * | 2017-11-02 | 2018-03-02 | 郑州云海信息技术有限公司 | A kind of GPU casees |
CN108173735A (en) * | 2018-01-17 | 2018-06-15 | 郑州云海信息技术有限公司 | A kind of GPU Box server cascaded communication method, apparatus and system |
CN108319539A (en) * | 2018-02-28 | 2018-07-24 | 郑州云海信息技术有限公司 | A kind of method and system generating GPU card slot position information |
CN108710418A (en) * | 2018-08-06 | 2018-10-26 | 郑州云海信息技术有限公司 | A kind of GPU-Switch structures cabinet |
CN108874726A (en) * | 2018-05-25 | 2018-11-23 | 郑州云海信息技术有限公司 | A kind of GPU whole machine cabinet PCIE link interacted system and method |
CN108959165A (en) * | 2018-06-28 | 2018-12-07 | 郑州云海信息技术有限公司 | A kind of management system of GPU whole machine cabinet cluster |
CN109189347A (en) * | 2018-09-20 | 2019-01-11 | 郑州云海信息技术有限公司 | A kind of sharing storage module, server and system |
CN109408440A (en) * | 2018-11-06 | 2019-03-01 | 郑州云海信息技术有限公司 | A kind of PCIE expanding unit |
CN110413557A (en) * | 2019-06-29 | 2019-11-05 | 苏州浪潮智能科技有限公司 | A kind of GPU accelerator |
TWI690789B (en) * | 2018-11-28 | 2020-04-11 | 英業達股份有限公司 | Graphic processor system |
CN111352494A (en) * | 2020-02-22 | 2020-06-30 | 苏州浪潮智能科技有限公司 | 54V input PCIE (peripheral component interface express) switch board power supply framework and power supply wiring method |
TWI704463B (en) * | 2019-03-29 | 2020-09-11 | 英業達股份有限公司 | Server system and management method thereto |
CN111736915A (en) * | 2020-06-05 | 2020-10-02 | 浪潮电子信息产业股份有限公司 | Management method, device, equipment and medium for cloud host instance hardware acceleration equipment |
CN114500413A (en) * | 2021-12-17 | 2022-05-13 | 阿里巴巴(中国)有限公司 | Equipment connection method and device and equipment connection chip |
Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20090251476A1 (en) * | 2008-04-04 | 2009-10-08 | Via Technologies, Inc. | Constant Buffering for a Computational Core of a Programmable Graphics Processing Unit |
CN104202194A (en) * | 2014-09-10 | 2014-12-10 | 华为技术有限公司 | Configuration method and device of PCIe (peripheral component interface express) topology |
CN104331130A (en) * | 2014-08-13 | 2015-02-04 | 浪潮电子信息产业股份有限公司 | A system based on a whole cabinet server ultra-large-scale deployment |
CN104915917A (en) * | 2015-06-01 | 2015-09-16 | 浪潮电子信息产业股份有限公司 | GPU cabinet, PCIe exchange device and server system |
CN105227666A (en) * | 2015-10-12 | 2016-01-06 | 浪潮(北京)电子信息产业有限公司 | The whole machine cabinet management framework that a kind of facing cloud calculates |
CN105426286A (en) * | 2015-11-05 | 2016-03-23 | 浪潮(北京)电子信息产业有限公司 | System for monitoring whole rack server |
CN106445045A (en) * | 2016-08-31 | 2017-02-22 | 浪潮电子信息产业股份有限公司 | Power supply copper bar and server |
CN106685725A (en) * | 2017-01-11 | 2017-05-17 | 郑州云海信息技术有限公司 | Central management control panel, method and system |
CN106774752A (en) * | 2017-01-11 | 2017-05-31 | 郑州云海信息技术有限公司 | A kind of Rack servers spare fans control method |
-
2017
- 2017-06-09 CN CN201710433600.0A patent/CN107239346A/en active Pending
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20090251476A1 (en) * | 2008-04-04 | 2009-10-08 | Via Technologies, Inc. | Constant Buffering for a Computational Core of a Programmable Graphics Processing Unit |
CN104331130A (en) * | 2014-08-13 | 2015-02-04 | 浪潮电子信息产业股份有限公司 | A system based on a whole cabinet server ultra-large-scale deployment |
CN104202194A (en) * | 2014-09-10 | 2014-12-10 | 华为技术有限公司 | Configuration method and device of PCIe (peripheral component interface express) topology |
CN104915917A (en) * | 2015-06-01 | 2015-09-16 | 浪潮电子信息产业股份有限公司 | GPU cabinet, PCIe exchange device and server system |
CN105227666A (en) * | 2015-10-12 | 2016-01-06 | 浪潮(北京)电子信息产业有限公司 | The whole machine cabinet management framework that a kind of facing cloud calculates |
CN105426286A (en) * | 2015-11-05 | 2016-03-23 | 浪潮(北京)电子信息产业有限公司 | System for monitoring whole rack server |
CN106445045A (en) * | 2016-08-31 | 2017-02-22 | 浪潮电子信息产业股份有限公司 | Power supply copper bar and server |
CN106685725A (en) * | 2017-01-11 | 2017-05-17 | 郑州云海信息技术有限公司 | Central management control panel, method and system |
CN106774752A (en) * | 2017-01-11 | 2017-05-31 | 郑州云海信息技术有限公司 | A kind of Rack servers spare fans control method |
Non-Patent Citations (1)
Title |
---|
厂商: "浪潮发布业界最高GPU密度的SR-AI整机柜", 《中关村在线》 * |
Cited By (24)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107748726A (en) * | 2017-11-02 | 2018-03-02 | 郑州云海信息技术有限公司 | A kind of GPU casees |
CN107748726B (en) * | 2017-11-02 | 2020-03-24 | 郑州云海信息技术有限公司 | GPU (graphics processing Unit) box |
CN108173735A (en) * | 2018-01-17 | 2018-06-15 | 郑州云海信息技术有限公司 | A kind of GPU Box server cascaded communication method, apparatus and system |
US11641405B2 (en) | 2018-01-17 | 2023-05-02 | Zhengzhou Yunhai Information Technology Co., Ltd. | GPU box server cascade communication method, device, and system |
CN108173735B (en) * | 2018-01-17 | 2020-08-25 | 苏州浪潮智能科技有限公司 | GPU Box server cascade communication method, device and system |
WO2019140921A1 (en) * | 2018-01-17 | 2019-07-25 | 郑州云海信息技术有限公司 | Gpu box server cascade communication method, device, and system |
WO2019165773A1 (en) * | 2018-02-28 | 2019-09-06 | 郑州云海信息技术有限公司 | Method and system for generating gpu card slot position information |
CN108319539A (en) * | 2018-02-28 | 2018-07-24 | 郑州云海信息技术有限公司 | A kind of method and system generating GPU card slot position information |
CN108319539B (en) * | 2018-02-28 | 2022-03-22 | 郑州云海信息技术有限公司 | Method and system for generating GPU card slot position information |
CN108874726A (en) * | 2018-05-25 | 2018-11-23 | 郑州云海信息技术有限公司 | A kind of GPU whole machine cabinet PCIE link interacted system and method |
CN108959165A (en) * | 2018-06-28 | 2018-12-07 | 郑州云海信息技术有限公司 | A kind of management system of GPU whole machine cabinet cluster |
CN108710418B (en) * | 2018-08-06 | 2023-09-22 | 郑州云海信息技术有限公司 | GPU-Switch structure case |
CN108710418A (en) * | 2018-08-06 | 2018-10-26 | 郑州云海信息技术有限公司 | A kind of GPU-Switch structures cabinet |
CN109189347A (en) * | 2018-09-20 | 2019-01-11 | 郑州云海信息技术有限公司 | A kind of sharing storage module, server and system |
CN109408440A (en) * | 2018-11-06 | 2019-03-01 | 郑州云海信息技术有限公司 | A kind of PCIE expanding unit |
TWI690789B (en) * | 2018-11-28 | 2020-04-11 | 英業達股份有限公司 | Graphic processor system |
TWI704463B (en) * | 2019-03-29 | 2020-09-11 | 英業達股份有限公司 | Server system and management method thereto |
CN110413557B (en) * | 2019-06-29 | 2020-11-10 | 苏州浪潮智能科技有限公司 | GPU (graphics processing unit) accelerating device |
CN110413557A (en) * | 2019-06-29 | 2019-11-05 | 苏州浪潮智能科技有限公司 | A kind of GPU accelerator |
CN111352494A (en) * | 2020-02-22 | 2020-06-30 | 苏州浪潮智能科技有限公司 | 54V input PCIE (peripheral component interface express) switch board power supply framework and power supply wiring method |
CN111736915A (en) * | 2020-06-05 | 2020-10-02 | 浪潮电子信息产业股份有限公司 | Management method, device, equipment and medium for cloud host instance hardware acceleration equipment |
CN111736915B (en) * | 2020-06-05 | 2022-07-05 | 浪潮电子信息产业股份有限公司 | Management method, device, equipment and medium for cloud host instance hardware acceleration equipment |
CN114500413A (en) * | 2021-12-17 | 2022-05-13 | 阿里巴巴(中国)有限公司 | Equipment connection method and device and equipment connection chip |
CN114500413B (en) * | 2021-12-17 | 2024-04-16 | 阿里巴巴(中国)有限公司 | Device connection method and device, and device connection chip |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107239346A (en) | A kind of whole machine cabinet computing resource tank node and computing resource pond framework | |
CN105549460A (en) | Satellite-borne electronic equipment comprehensive management and control system | |
CN101819556B (en) | Signal-processing board | |
CN107659437A (en) | A kind of whole machine cabinet computing resource Pooled resources automatic recognition system and method | |
CN104135514B (en) | Fusion type virtual storage system | |
CN103336756B (en) | A kind of generating apparatus of data computational node | |
CN107480094A (en) | A kind of pond server system architecture of fusion architecture | |
CN105159617A (en) | Pooled storage system framework | |
CN102402474B (en) | Prototype verification device for programmable logic devices | |
CN104641593B (en) | Web plate and communication equipment | |
CN104579786B (en) | A kind of server design method based on 2D Torus network topology architectures | |
CN203178870U (en) | Internet access switching card | |
CN104750581A (en) | Redundant interconnection memory-shared server system | |
CN210983137U (en) | Server hardware system architecture | |
CN102298418A (en) | Advanced mezzanine card (AMC) board card based on MicroTCA standard and connection method thereof | |
CN209248518U (en) | A kind of solid state hard disk expansion board clamping and server | |
CN207440541U (en) | A kind of redundancy communication controller based on arm processor | |
CN102495819B (en) | Method for realizing blade service high speed bus SI (System Information) optimization and redundancy through one-third orthogonal intersection | |
CN204928853U (en) | Simple and easy serial communication equipment | |
CN206877374U (en) | A kind of power marketing intelligent platform | |
CN106484656B (en) | A kind of management board of collectable multinode management information | |
Miyoshi et al. | New system architecture for next-generation green data centers: mangrove | |
CN109739560A (en) | A kind of GPU card cluster configuration control system and method | |
CN105468104A (en) | Converged server and backboard | |
CN209401013U (en) | A kind of server and its memory Riser plate |
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 |
Application publication date: 20171010 |
|
RJ01 | Rejection of invention patent application after publication |