CN104679714A - Supercomputer cluster based on ATCA (advanced telecom computing architecture) - Google Patents
Supercomputer cluster based on ATCA (advanced telecom computing architecture) Download PDFInfo
- Publication number
- CN104679714A CN104679714A CN201510103736.6A CN201510103736A CN104679714A CN 104679714 A CN104679714 A CN 104679714A CN 201510103736 A CN201510103736 A CN 201510103736A CN 104679714 A CN104679714 A CN 104679714A
- Authority
- CN
- China
- Prior art keywords
- blade
- atca
- supercomputer
- exchange
- cluster
- 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
Abstract
The invention discloses a supercomputer cluster based on an ATCA (advanced telecom computing architecture). A computing blade in an ATCA host case is connected with a switch blade through an ATCA bus, the switch blade is connected with a core switch, and the core switch is interconnected with a management server. The supercomputer cluster has the advantages that by adopting the computing node reconfiguration scheme, the scale of the cluster can be flexibly determined, the different computing nodes can be adopted according to requirements, a high-property computing cluster for different special fields is formed, a supercomputer platform with wide purpose is provided, the price-performance ratio of the supercomputer cluster is improved, and the potential user group is increased.
Description
Technical field
the present invention relates to a kind of supercomputer, relate in particular to one and utilize ATCA framework to realize high density, small size, clustered node, multiduty high-performance supercomputer cluster.
Background technology
aTCA(Advanced Telecom Computing Architecture) the standard i.e. telecommunications computing platform of advanced person, it is derived from the widely used mainstream industry computing technique of new generation in field---CompactPCI standards such as telecommunications, space flight, Industry Control, medicine equipment, intelligent transportation, military equipments.The high performance-price ratio provided for converged communication of future generation and data network application, based on modular construction, compatible and extendible hardware architecture.
supercomputer is the computer that can perform large data quantity that general PC cannot process and high-speed computation.The concept of its basic composition assembly and PC is without too big-difference, but specification and performance are then powerful many, are a kind of ultra-large type robot calculator.Have the ability of very strong calculating and process data, principal feature shows as high speed and Large Copacity, is furnished with multiple outside and peripherals and abundant, H.D software systems.Supercomputer is the class computing machine that in computing machine, function is the strongest, arithmetic speed is the fastest, memory capacity is maximum, be used for national high-tech area and sophisticated technology research, be the embodiment of a national research strength, it is to national security, and economy and social development have very important meaning.It is the important symbol of national science and technology development level and overall national strength.
ethernet (Ethernet) refers to and is created and the baseband LAN specification developed jointly by Xerox, Intel and DEC by Xerox company, is the most general communication protocol standard that current existing LAN (Local Area Network) adopts.Ethernet uses CSMA/CD(csma and collision detection) technology, and operate on polytype cable with the speed of 10M/S.Ethernet and IEEE802.3 series standard similar.Comprise the Ethernet (10Mbit/s) of standard, Fast Ethernet (100Mbit/s) and 10G(10Gbit/s) Ethernet.They all meet IEEE802.3.
optical fiber network interface card, fiber optic Ethernet network interface card or fiber optic Ethernet adapter.Host-host protocol is TCP/IP Ethernet protocol, is generally connected with fibre channel media by optical fiber cable.Interface type is divided into light mouth and electric mouth.Light mouth is all generally carry out data transmission by optical fiber cable, and interface module is generally SFP(transfer rate 2Gb/s) and GBIC(1Gb/s), corresponding interface is SC and LC.The interface type of electricity mouth is generally DB9 pin or HSSDC.
high-performance calculation (High Perfermance Computing) group of planes, is called for short a HPC group of planes.A this kind of group of planes mainly solves the calculating of extensive problem in science and the process of mass data, as scientific research, weather forecast, calculating simulation, military field engineering, CFD/CAE, bio-pharmaceuticals, gene sequencing, image procossing etc.
network exchange refers to by certain equipment, as switch etc., different signals or signal form is converted to the discernible signal type of the other side thus reaches communication objective a kind of exchanging form, common are: exchanges data, circuit switched, MESSAGE EXCHANGE, packet switch.
ethernet switch is the switch based on Ethernet transmission data, and Ethernet adopts the LAN (Local Area Network) of shared-bus transmission medium mode.To be that each port is direct be connected with main frame the structure of Ethernet switch, and be generally all operated in full duplex mode.Exchange function is communicated with many to port simultaneously, makes every main frame mutually communicated for a pair as exclusive communication medium, can carry out ensuring escapement from confliction transmission data.
server, also claims servomechanism.Server is the high-performance computer in network environment, and it intercepts the services request that other computing machines (client computer) on network are submitted to, and provides corresponding service, and for this reason, server must have the service born and ensure the ability of service.
iIC and Inter-Integrated Circuit (IC bus), this bus type is designed at early eighties by Philips Semiconductor Co., Ltd., mainly be used for connecting overall circuit (ICS), IIC is a kind of multidirectional control bus, that is, under multiple chip can be connected to same bus structure, each chip can as the control source of real-time Data Transmission simultaneously.This mode simplifies signal transmission bus interface.
existing supercomputer cluster often volume is very huge, and density is not high, and the scale of node and use etc. are all very limited.
Summary of the invention
in order to overcome the deficiencies in the prior art, the object of the invention is: a kind of supercomputer cluster based on ATCA framework is provided, volume is little, novel structure, compact, density is high, by Ethernet to all computing nodes and management server interconnected, management server is managed all nodes by network, node scalable, use flexibly, control mode is ripe, system stability.
technical scheme of the present invention is:
based on a supercomputer cluster for ATCA framework, each blade that calculates is connected with exchange blade by ATCA bus in ATCA cabinet, and described exchange blade is connected with core switch, described core switch and management server interconnected.
preferably, described calculating blade is exchanged by multistage network and converges.
preferably, described calculating blade connects computing node bar by connector.
preferably, described exchange blade is at least one piece.
preferably, described exchange blade comprises gigabit PHY, 10,000,000,000 PHY, exchange chip and managing chip, and described gigabit PHY receives and converges from calculating blade the network signal of coming; Described 10,000,000,000 PHY are the passage with ATCA cabinet external communication; The exchanges data that described exchange chip realizes 24 gigabit networkings and 2 10,000,000,000 networks is interconnected; Described managing chip is in charge of control and configuration information.
preferably, described calculating blade comprises node bar interface, administration module, Switching Module and power module; Described power module is powered by the 48V of ATCA bus standard, described calculating blade is provided with IIC interface, Shelf Management Module carries out communication by iic bus and administration module, and described Switching Module is node and the communication channel calculated outside blade, connects simultaneously and calculates each computing node of inner blade.
in ATCA cabinet of the present invention, each blade that calculates is connected with exchange blade by ATCA bus, thus realizes the interconnected of network; ATCA bus is the bus structure of standard universal, stable performance, compatible strong; Meanwhile, ATCA bus supports that the calculating blade of variable number works simultaneously, and for the position calculating blade and the number of nodes no standard calculated on blade, applying flexible, can according to the scale of different purposes Adjustable calculation clusters and allocation plan.Calculate blade adopt memory bar with connector connects computing node bar, stable performance, use flexible, the node bar of difference in functionality can be changed as required, save the resource of computing cluster, extend the range of application of computing cluster simultaneously, increase substantially the cost performance of computing cluster.Adopt 10,000,000,000 network switchs interconnected between ATCA cabinet and management server, transport tape is roomy, and extensibility is strong, adopts ripe network technology transmission performance is stablized and is easy to management maintenance.Management board can carry out point-to-point management by iic bus to each blade and node, can carry out maintenance management to an independent part, also can batch maintenance management, improve the work efficiency of computing cluster greatly, reduce the maintenance cost of cluster.
advantage of the present invention is:
1. present invention employs computing node restructural scheme, the scale of cluster can be determined flexibly, also different computing nodes can be adopted according to demand, not only can become the HPCC in different special field, also provide a broad-spectrum super calculation platform simultaneously, improve the super cost performance calculating cluster, add potential user group.
2. novel structure of the present invention, compact, density is high, can low-power consumption, realize the high-effect calculating of different application at low cost, wide accommodation, stable performance, safe and reliable.
3. the present invention adopts ATCA framework, and volume is little, and density is high, the highest configurable 192 computing nodes of ATCA cabinet inside of a 14U, and volume is only equivalent to 3 ~ 4 Small Universal servers; The present invention adopts the full Ethernet communication of cluster internal, and transmission performance is stablized, and be easy to management, development difficulty is low, applying flexible; Present invention employs connector and connect conveniently replaced computing node bar, can change different computing node according to different purposes, while maintenance high-performance calculation, range of application is more extensive, and application mode is more flexible.
Accompanying drawing explanation
below in conjunction with drawings and Examples, the invention will be further described:
fig. 1 is the overall network Organization Chart of the supercomputer cluster that the present invention is based on ATCA framework;
fig. 2 is the layered network architecture figure of the supercomputer cluster that the present invention is based on ATCA framework;
fig. 3 is the ATCA casing structure schematic diagram of the supercomputer cluster that the present invention is based on ATCA framework;
fig. 4 is the exchange blade structure block diagram of the supercomputer cluster that the present invention is based on ATCA framework;
fig. 5 is the calculating blade structure block diagram of the supercomputer cluster that the present invention is based on ATCA framework.
Embodiment
for making the object, technical solutions and advantages of the present invention clearly understand, below in conjunction with embodiment also with reference to accompanying drawing, the present invention is described in more detail.Should be appreciated that, these describe just exemplary, and do not really want to limit the scope of the invention.In addition, in the following description, the description to known features and technology is eliminated, to avoid unnecessarily obscuring concept of the present invention.
embodiment:
below in conjunction with concrete drawings and Examples, the invention will be further described.
as shown in Figure 1: in this example, have 2048 computing nodes, they are through the continuous convergence of exchange network, finally be connected on 10,000,000,000 switches, management server, by connecting this switch, can carry out communication with each computing node in cluster easily.
as shown in Figure 2: as shown in network stratified structure, Access Layer is on calculating blade, and the exchange network of each bottom connects four node bars, and each node bar has two computing nodes in this example, so, this exchange network has converged the network data of 8 nodes.Meanwhile, the two grade network calculating blade exchanges and has converged the elementary network of two-way, so the interface exchanged by two grade network, just can carry out network communication by 16 computing nodes on the calculating blade therewith in example.Exchange blade is Network Convergence Layer, the network of each calculating blade converges in the aggregation networks exchanging blade, adopts the ATCA cabinet of 14U in this example, and configurable at most 12 pieces calculate blade, namely in this example, an exchange blade can converge at most 192 computing nodes.Finally, the network of convergence-level is all connected to core exchange layer, and management server carries out communication by being connected into core exchange layer with all nodes, and the computing node quantity configured in this example is 2048.
as shown in Figure 3: the backboard of ATCA cabinet leaves the slot of 14 blades, plan that 12 calculate blade, 1 exchange blade and 1 exchange blade for subsequent use in this example.The position of their distributions as shown in the figure.In use engineering, at least one piece must be had to exchange blade, complete networks converge function, and calculate quantity and the insertion position of blade, can determine as the case may be.
as shown in Figure 4: the aggregation networks exchanging blade is made up of four parts: comprise gigabit PHY, 10,000,000,000 PHY, exchange chip and managing chip.Gigabit PHY receives and converges from calculating blade the network signal of coming; 10000000000 PHY are the passages with ATCA cabinet external communication; Exchange chip realizes the exchanges data interconnecting function of 24 gigabit networkings and 2 10,000,000,000 networks; Managing chip is in charge of control and the configuration information of first three part.By the combination of these four parts, exchange blade and not only achieve the different communication calculated between blade, also achieve the communication calculating blade and ATCA cabinet outside simultaneously.Meanwhile, the employing of 10,000,000,000 nets also maintains the bandwidth of each calculating blade to greatest extent.
as shown in Figure 5: calculating the power supply of blade is that the 48V of ATCA bus standard powers, and can reduce the power loss in transmitting procedure like this.Blade has an IIC interface, Shelf Management Module can carry out communication by iic bus and administration module, thus the power supply of Management Calculation blade and computing node configuration network parameter.Calculating most important ingredient on blade is the interface of 8 node bars, and can configure different node bars herein, in this example, each node bar comprises two computing nodes.The Switching Module calculating blade is node and the communication channel calculated outside blade, also achieves the interconnected of each computing node of inner blade simultaneously.With upper part, jointly constitute calculating blade, become the carrier of computing node in system.
the present invention adopts ATCA framework, and adopts network communications technology to carry out data transmission and computing node management.ATCA framework is a ripe structure system, the invention this framework and super cluster of calculating are combined, and provides that one stable, volume is little, density is high, flexible function, and consume energy few super calculation cluster.
present invention employs computing node restructural scheme, the scale of cluster can be determined flexibly, also different computing nodes can be adopted according to demand, not only can become the HPCC in different special field, also provide a broad-spectrum super calculation platform simultaneously, improve the super cost performance calculating cluster, add potential user group.
novel structure of the present invention, compact, density is high, can low-power consumption, realize the high-effect calculating of different application at low cost, wide accommodation, stable performance, safe and reliable.
should be understood that, above-mentioned embodiment of the present invention only for exemplary illustration or explain principle of the present invention, and is not construed as limiting the invention.Therefore, any amendment made when without departing from the spirit and scope of the present invention, equivalent replacement, improvement etc., all should be included within protection scope of the present invention.In addition, claims of the present invention be intended to contain fall into claims scope and border or this scope and border equivalents in whole change and modification.
Claims (6)
1. based on a supercomputer cluster for ATCA framework, it is characterized in that, each blade that calculates is connected with exchange blade by ATCA bus in ATCA cabinet, and described exchange blade is connected with core switch, described core switch and management server interconnected.
2. the supercomputer cluster based on ATCA framework according to claim 1, is characterized in that, described calculating blade is exchanged by multistage network and converges.
3. the supercomputer cluster based on ATCA framework according to claim 1 and 2, is characterized in that, described calculating blade connects computing node bar by connector.
4. the supercomputer cluster based on ATCA framework according to claim 1, is characterized in that, described exchange blade is at least one piece.
5. the supercomputer cluster based on ATCA framework according to claim 1, is characterized in that, described exchange blade comprises gigabit PHY, 10,000,000,000 PHY, exchange chip and managing chip, and described gigabit PHY receives and converges from calculating blade the network signal of coming; Described 10,000,000,000 PHY are the passage with ATCA cabinet external communication; The exchanges data that described exchange chip realizes 24 gigabit networkings and 2 10,000,000,000 networks is interconnected; Described managing chip is in charge of control and configuration information.
6. the supercomputer cluster based on ATCA framework according to claim 1, is characterized in that, described calculating blade comprises node bar interface, administration module, Switching Module and power module; Described power module is powered by the 48V of ATCA bus standard, described calculating blade is provided with IIC interface, Shelf Management Module carries out communication by iic bus and administration module, and described Switching Module is node and the communication channel calculated outside blade, connects simultaneously and calculates each computing node of inner blade.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510103736.6A CN104679714A (en) | 2015-03-10 | 2015-03-10 | Supercomputer cluster based on ATCA (advanced telecom computing architecture) |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510103736.6A CN104679714A (en) | 2015-03-10 | 2015-03-10 | Supercomputer cluster based on ATCA (advanced telecom computing architecture) |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN104679714A true CN104679714A (en) | 2015-06-03 |
Family
ID=53314781
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510103736.6A Pending CN104679714A (en) | 2015-03-10 | 2015-03-10 | Supercomputer cluster based on ATCA (advanced telecom computing architecture) |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104679714A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105117364A (en) * | 2015-09-18 | 2015-12-02 | 江苏微锐超算科技有限公司 | Service calculation blade based on ATCA |
| CN106648900A (en) * | 2016-12-28 | 2017-05-10 | 深圳Tcl数字技术有限公司 | Smart television-based supercomputing method and system |
| CN109885447A (en) * | 2018-12-27 | 2019-06-14 | 曙光信息产业(北京)有限公司 | The detecting and management system of clustered node |
| CN110769391A (en) * | 2019-10-22 | 2020-02-07 | 玉溪市昊协科技有限公司 | Environmental technology warehouse communication system |
| CN111008174A (en) * | 2019-12-06 | 2020-04-14 | 深圳市时代通信技术有限公司 | ATCA-based 100GE high-density server system |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101150413A (en) * | 2007-10-31 | 2008-03-26 | 中兴通讯股份有限公司 | A multi-frame cascading system and method for ATCA knife server |
| CN101557294A (en) * | 2009-05-08 | 2009-10-14 | 中兴通讯股份有限公司 | Machine frame cascading networking system of ATCA blade server and method thereof |
| CN101895444A (en) * | 2010-07-28 | 2010-11-24 | 南京信息工程大学 | Dual system of ATCA blade server, connection method and test method |
| CN103428114A (en) * | 2013-08-08 | 2013-12-04 | 曙光信息产业股份有限公司 | ATCA (advanced telecom computing architecture) 10-gigabit switching board and system |
| EP2709318A1 (en) * | 2012-09-13 | 2014-03-19 | Alcatel Lucent | Blade and Advanced Mezzanine Card AMC Activation Control In An Advanced Telecom Computing Architecture ATCA Shelf |
| CN103984390A (en) * | 2014-05-22 | 2014-08-13 | 华为技术有限公司 | Blade and blade server |
-
2015
- 2015-03-10 CN CN201510103736.6A patent/CN104679714A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101150413A (en) * | 2007-10-31 | 2008-03-26 | 中兴通讯股份有限公司 | A multi-frame cascading system and method for ATCA knife server |
| CN101557294A (en) * | 2009-05-08 | 2009-10-14 | 中兴通讯股份有限公司 | Machine frame cascading networking system of ATCA blade server and method thereof |
| CN101895444A (en) * | 2010-07-28 | 2010-11-24 | 南京信息工程大学 | Dual system of ATCA blade server, connection method and test method |
| EP2709318A1 (en) * | 2012-09-13 | 2014-03-19 | Alcatel Lucent | Blade and Advanced Mezzanine Card AMC Activation Control In An Advanced Telecom Computing Architecture ATCA Shelf |
| CN103428114A (en) * | 2013-08-08 | 2013-12-04 | 曙光信息产业股份有限公司 | ATCA (advanced telecom computing architecture) 10-gigabit switching board and system |
| CN103984390A (en) * | 2014-05-22 | 2014-08-13 | 华为技术有限公司 | Blade and blade server |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105117364A (en) * | 2015-09-18 | 2015-12-02 | 江苏微锐超算科技有限公司 | Service calculation blade based on ATCA |
| CN106648900A (en) * | 2016-12-28 | 2017-05-10 | 深圳Tcl数字技术有限公司 | Smart television-based supercomputing method and system |
| CN106648900B (en) * | 2016-12-28 | 2020-12-08 | 深圳Tcl数字技术有限公司 | Supercomputing method and system based on smart television |
| CN109885447A (en) * | 2018-12-27 | 2019-06-14 | 曙光信息产业(北京)有限公司 | The detecting and management system of clustered node |
| CN110769391A (en) * | 2019-10-22 | 2020-02-07 | 玉溪市昊协科技有限公司 | Environmental technology warehouse communication system |
| CN111008174A (en) * | 2019-12-06 | 2020-04-14 | 深圳市时代通信技术有限公司 | ATCA-based 100GE high-density server system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| AU2018200158B2 (en) | Method And Apparatus To Manage The Direct Interconnect Switch Wiring And Growth In Computer Networks | |
| US20130156425A1 (en) | Optical Network for Cluster Computing | |
| CN104679714A (en) | Supercomputer cluster based on ATCA (advanced telecom computing architecture) | |
| CN206820773U (en) | A kind of board for supporting RapidIO and network double crossing over function | |
| Schares et al. | Optics in future data center networks | |
| CN204650513U (en) | Distributed structure/architecture equipment and serial port circuit thereof | |
| CN105099776A (en) | Cloud server management system | |
| CN104035525A (en) | Computational node | |
| CN103116559A (en) | Design method for high-speed interconnecting server system | |
| CN205304857U (en) | 10, 000, 000, 000 light network switch | |
| Vahdat | Delivering scale out data center networking with optics—Why and how | |
| CN103605413A (en) | Rack-mounted server system cabinet, rack-mounted server system and management method thereof | |
| CN208046650U (en) | A kind of hybrid switching board | |
| CN101026575A (en) | High available, high scalble modularized network server system structure | |
| CN103716258B (en) | High-density line card, switching device, cluster system and electric signal type configuration method | |
| CN105577752A (en) | Management system used for fusion framework server | |
| CN214799500U (en) | Network adapter | |
| Minkenberg et al. | Large-scale system partitioning using OCS | |
| CN202720644U (en) | Graphic processing unit (GPU) high-performance computing platform device applied to electromagnetic simulation | |
| CN115996204B (en) | Out-of-band Ethernet interface switching device, multi-node server system and server equipment | |
| CN208158628U (en) | Police's jaws equipment and system for the acquisition of information security data | |
| CN208209997U (en) | A kind of data transmission system | |
| Shao et al. | OeIM: An Optoelectronic Interconnection Middleware for the Exascale Computer | |
| CN116962220A (en) | Full-dimension definable intelligent communication network device | |
| Yoo | The role of photonics in future computing systems and data centers |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into 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: 20150603 |