CN1456994A - Self-organizing dynamic network computer system structure - Google Patents
Self-organizing dynamic network computer system structure Download PDFInfo
- Publication number
- CN1456994A CN1456994A CN 03142336 CN03142336A CN1456994A CN 1456994 A CN1456994 A CN 1456994A CN 03142336 CN03142336 CN 03142336 CN 03142336 A CN03142336 A CN 03142336A CN 1456994 A CN1456994 A CN 1456994A
- Authority
- CN
- China
- Prior art keywords
- gridding
- computer system
- organization
- computer
- dynamic
- 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.)
- Granted
Links
Images
Landscapes
- Multi Processors (AREA)
Abstract
A computer system organization is a new computer pattern which automatically utilizes the independent and gridded functional component service in high-speed network to organize a computer system and to operate application program effectively according to the application requirements of computer pattern and operation pattern. It also comprises high-speed exchange equipment connected with two CPU frameworks, dynamic random storage framework and magnetic disc framework separately for being used at optical exchange message.
Description
Technical field
The present invention relates to the grid computing technology field, particularly a kind of dynamic self-organization Computer Architecture of gridding of novel high-performance computer system.
Background technology
It is an important directions of Computer Architecture development in recent years that grid (grid) calculates.But the research of present grid computing also mainly is in the junior stage, and promptly carries out resource sharing by the internet and original resource isolated island is eliminated in collaborative work.The elementary cell that constitutes grid system is computer system independently one by one still, and full-order system of each computer system formation itself has complete operating system and application software.Computing system then comes collaborative work by one deck application layer services software that is structured on the local computer system, belongs to sharing of a kind of coarseness.
On the other hand, the develop rapidly of network interconnection technique such as optical interconnection in recent years, the optical fiber interconnections technology between the equipment has begun to use, and based on the short distance interconnection technique of optical waveguide, the free space optical interconnection technology, the exchange system of full light etc. also develops rapidly.The optical interconnection of estimating following 2 ~ 5 years computing machine plate levels can move to maturity, and the optical interconnection of chip-scale also might become a reality in 5 ~ 10 years.Therefore, light connects the trend that supercomputer system inside is a certainty that enters.
The characteristics of optical interconnection are to support long Distance Transmission of data and high-transmission bandwidth, and this just provides possibility for the high-speed data exchange between the calculating unit on the greater room yardstick.Can be comparable on the bandwidth theory that the employing optical interconnection network connects with the interconnected bandwidth of inside computer system, greatly reduce the difference between local resource and the remote resource.This will drive the more fine-grained shared model of gridding technique trend, thereby strides forward to a brand-new stage.
In this new developing stage, the unit that constitutes the gridding computer system will no longer be complete one by one computer system, but independently, the functional part of gridding, as network C PU, network RAM, network disk or the like.Grid will be not only " computing machine couples together by network ", but become truly " computing machine that network couples together ".
Summary of the invention
In view of this, an object of the present invention is by functional part long-range on the express network is that the functional part of gridding makes up the gridding computer system.Every kind of functional part all forms a kind of network service that independently has the self-description function, and computer system then constitutes a comprehensive total system by parts services different on the network by the express network interconnection.Accompanying drawing 1 provides an example.
Because available AFU service at present has only the network disk memory device, another object of the present invention is to propose independently CPU (central processing unit) (CPU) service and dynamic RAM (DRAM) service.After adopting functional part clastotypes such as CPU and DRAM, will be convenient to the high density industrial design of each parts, and conveniently share, subregion and fault-tolerance design.
Because the grid function component distribution is different local network, will no longer have computer system fixing, physics on the traditional sense.Because the resource of traditional high-performance computer system just immobilizes after manufacturing and designing usually, the functional part that an object of the present invention is to constitute a computer system can and discharge by the dynamic application of network, allocates dynamically according to demands of applications and available resource.Local unit failure does not have influence on the normal operation of overall applicability.The one-piece construction of system also can dynamically change when needing.The dynamic-configuration change of this resource is not limited to existing internal system, but comprises any functional unit service that is distributed on the network.
Resource distribution in traditional computer system is fixed, and the design of computer system coordinates to constitute jointly computer system based on predetermined resource by fixing steering logic, firmware and system software.Because the default system of this resource lacks enough dirigibilities, one object of the present invention is that the computer system that proposes a kind of self-organization makes up pattern, be characterized in utilizing the functional unit of gridding to have the function of self-description, automatically the functional unit service that search satisfies the demands on network, automatically carry out resources allocation and negotiation, collect enough computational resources and set up association between functional part automatically, to be organized into a complete computer system.
Traditional application program normally is optimized design at the variation of fixing Computer Architecture.And the variation of the model of various science computational problems and algorithm be relative slowly, but the variation of Computer Architecture is but very fast relatively.Because the situation of this traditional " application adaptation architecture ", one object of the present invention is to adopt the functional part of gridding dynamically to make up the Computer Architecture that is adapted to application-specific, can construct similar symmetrical multiprocessing according to the needs of application model, a group of planes, the system of various structures such as single instruction stream multiple data stream and unitized construction.Thereby realize the computation schema of a kind of " Architecture On Demand " (architecture on-demand customization).
The objective of the invention is to realize by following technical scheme:
A kind of Computer Architecture of dynamic self-organization of gridding can be utilized the functional part service dynamic group of gridding in the network to be made into a computer system, and effectively run application automatically according to the computation schema of using and the needs of operational mode.
Difference in functionality parts in this computer system are not what to concentrate physically, but are distributed on the network, and the share service of Internet resources is not to be unit with the complete machine, but is unit with parts.
The functional part service mode of gridding, CPU (central processing unit) and internal memory are separated, independently CPU (central processing unit) service and internal memory service are provided, make internal memory to share for all CPU (central processing unit) easily or to divide, eliminate the limitation of conventional central processing unit and local internal memory binding.
Constituting the functional unit quantity of computer system and the syndeton between the functional unit can dynamic change, the variation of demand when adapting to the application program operation.
The building process of computer system is self-organization rather than predefined, in the building process of system, inquire about dynamically, the available resources on negotiation and the dispatch network and set up related, computer system of self-organization formation.
The structure of computer system can be constructed neatly as required according to demands of applications, realizes various architectures and combinations thereof such as symmetrical multiprocessing, massively parallel processing, a group of planes, streamline, single instruction stream multiple data stream.
The dynamic self-organization mode can be applied to the various computer nodes on the existing the Internet too, the transformable the Internet of the dynamic structure level computing grid system that the partial function parts of the existing node of utilization or the part service function of node are formed.
A kind of gridding computer system comprises:
The high speed switching equipment is connected to two CPU frames, dynamic RAM frame and disk (DISK) machine; Be used for the light exchange message.
What one of them CPU frame adopted is to share consistent buffer memory symmetric multiprocessor structure, and another CPU frame then is based on the MPP structure of two-dimentional Mesh internet.
Though the invention belongs to a kind of novel Computer Architecture, its basic thought is adapted to the computer resource of existing all kinds of griddings on the current the Internet equally.Various computing machines have huge calculating potentiality on the Internet because have now, one object of the present invention is that the functional part of gridding also can be part of functions parts such as disk, internal memory or the part service function of the existing computer node on the Internet, as database, file system or the like, make full use of existing computational resource and support existing application, thereby can make up the Internet level computing grid system of dynamically can recombinating on a large scale.
Description of drawings
Fig. 1 is the gridding computer system;
Fig. 2 is gridding functional part server logic figure.
Among Fig. 1, " X " expression is based on the high speed switching equipment 1 of light switching technology.One of each box indicating is frame (Rack) independently, what one of them CPU frame adopted is to share consistent buffer memory symmetric multiprocessor structure 3, another CPU frame then is based on the MPP structure 2 of two-dimentional Mesh internet, and these CPU visit the high capacity internal memory of 5 li of DRAM (dynamic RAM) frames and the magnanimity disk storage service of disk (DISK) frame 4 the insides by optical internet.Frame by increasing different functional parts can also the dynamic expansion system scale and function.
1) on express network, sets up the service of gridding functional part
But the functional part service of gridding should have basic characteristics such as express network interface, self-described and management function partition functionality unit, has provided gridding functional part server logic figure as accompanying drawing 2.
Among Fig. 2, a gridding functional part server is made up of four parts:
adopt high speed network interface such as optical interconnection guarantee outside system can be directly, the visit of high speed
Ask the functional part that server is inner specific.
Service describing that is embedded and negotiation functionality are used for describing book server institute tool to external client
Service ability that has and available resources state, and according to certain agreement distribution idling-resource.
Management that is embedded and scheduling feature assurance function parts services device do not rely on and outside specific are
Just can independently, effectively there be and provides service in system.
The functional unit assurance resource that can divide can fully be shared and that by the different application in outside
Do not disturb mutually around here.
The service of gridding functional part can be through carrying out the novel highly integrated server of industrial design again, also can be the directly externally service of output of partial function module of active computer node.
2) by Intellective Management Software dynamic collection and the managerial grid parts of formatting
After the gridding parts services is set up, can use the needs that move thereby satisfy by intelligence software News Search, application, management, release gridding parts services.The gridding component distribution is in the network different location, but all has the self-described function.Intelligence software can be by some index service or is directly found the position of the gridding parts services device that may satisfy the demand, hold consultation with relevant gridding parts services device, after taking all factors into consideration factors such as performance, price, time, the combination of the optimum of application demand is satisfied in selection.This gridding parts collection process is not used operational process all the time but be applied in before not being only limited in and using initial start.The intelligence software here can be the part of application program, perhaps itself constitutes a kind of network service.
3) dynamically change the organizational form of computer system according to application demand
Different according to the difference of the computation schema of using and operation stages are organized the gridding functional part by different modes.By changing logical relation, the interconnection pattern between the functional unit, the computer system that application program is seen has different organizational forms and program operating mode, thereby carries out in the mode of optimum.Subordinate list 3 has provided the corresponding of some common concurrent computer architectures and organizational form.Can also further set up the various integrated modes of these basic organizational forms.The adjustment that changes organizational form and resource can be that the part of application program also can constitute a kind of network service.
The above only is some basic explanations under the technology of the present invention design, and according to any equivalent transformation that technical scheme of the present invention is done, all should belong to protection scope of the present invention.
Illustrate: { } symmetric relation,＜in proper order/binding relationship ,+high-speed interconnect
| Traditional architectures | The functional unit mapping |
| Symmetry multiprocessing (SMP) | ???{CPU} n+{Mem} m+{Disk} k |
| Massively parallel processing (MPP) | ???{<CPU，Mem>} n+{Disk} m |
| A group of planes (cluster) | ???{<CPU，Mem，Disk>} n |
| Streamline (pipeline) | ???<CPU> n+{Mem} m+{Disk} k |
| Single instruction stream multiple data stream (SIMD) | ???MCPU+<VCPU，Mem> n+{Disk} m |
The CPU CPU (central processing unit), Mem internal memory (DRAM)
The Disk disk, MCPU management/Control Node, VCPU virtual machine
Table 3
Claims (9)
1, a kind of Computer Architecture of dynamic self-organization of gridding, can be according to the computation schema of using and the needs of operational mode, automatically utilize the functional part service dynamic group of gridding in the network to be made into a computer system, and effectively run application.
2, the dynamic self-organization Computer Architecture of gridding as claimed in claim 1, it is characterized in that, difference in functionality parts in this computer system are not to concentrate physically, but be distributed on the network, the share service of Internet resources is not to be unit with the complete machine, but is unit with parts.
3, the Computer Architecture of the dynamic self-organization of gridding as claimed in claim 2, it is characterized in that, the functional part service mode of gridding, CPU (central processing unit) and internal memory are separated, independently CPU (central processing unit) service and internal memory service are provided, make internal memory to share for all CPU (central processing unit) easily or to divide, eliminate the limitation of conventional central processing unit and local internal memory binding.
4, the dynamic self-organization Computer Architecture of gridding as claimed in claim 1, it is characterized in that, constituting the functional unit quantity of computer system and the syndeton between the functional unit can dynamic change, the variation of demand when adapting to the application program operation.
5, the dynamic self-organization Computer Architecture of gridding as claimed in claim 1, it is characterized in that, the building process of computer system is self-organization rather than predefined, in the building process of system, inquire about dynamically, the available resources on negotiation and the dispatch network and set up related, computer system of self-organization formation.
6, the dynamic self-organization Computer Architecture of gridding as claimed in claim 1, it is characterized in that, the structure of computer system can be constructed neatly as required according to demands of applications, realizes various architectures and combinations thereof such as symmetrical multiprocessing, massively parallel processing, a group of planes, streamline, single instruction stream multiple data stream.
7, the dynamic self-organization Computer Architecture of gridding as claimed in claim 1, it is characterized in that, the dynamic self-organization mode can be applied to the various computer nodes on the existing the Internet too, the transformable the Internet of the dynamic structure level computing grid system that the partial function parts of the existing node of utilization or the part service function of node are formed.
8, a kind of gridding computer system comprises:
High speed switching equipment (1) is connected to two CPU frames (2), (3), dynamic RAM frame (5) and disk (DISK) frame (4); Be used for the light exchange message.
9, gridding computer as claimed in claim 8 system is characterized in that, what one of them CPU frame adopted is to share consistent buffer memory symmetric multiprocessor structure, and another CPU frame then is based on the MPP structure of two-dimentional Mesh internet.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNB031423361A CN100351829C (en) | 2003-06-13 | 2003-06-13 | Self-organizing dynamic network computer system structure |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNB031423361A CN100351829C (en) | 2003-06-13 | 2003-06-13 | Self-organizing dynamic network computer system structure |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN1456994A true CN1456994A (en) | 2003-11-19 |
| CN100351829C CN100351829C (en) | 2007-11-28 |
Family
ID=29411834
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CNB031423361A Expired - Fee Related CN100351829C (en) | 2003-06-13 | 2003-06-13 | Self-organizing dynamic network computer system structure |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN100351829C (en) |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN100391149C (en) * | 2005-05-12 | 2008-05-28 | 中国科学院计算技术研究所 | Application managing sub-system in multi-node intelligent network application service system and method |
| CN100428167C (en) * | 2004-03-19 | 2008-10-22 | 国际商业机器公司 | Computer system, server constituting the same, job execution control method thereof, and program |
| CN100435159C (en) * | 2006-08-24 | 2008-11-19 | 上海交通大学 | Method for implementing mesh based optical mesh emulation platform |
| CN100440832C (en) * | 2004-08-20 | 2008-12-03 | 清华大学 | Method for building self-organized network skeleton structure |
| CN101226518B (en) * | 2008-01-25 | 2010-11-03 | 华南理工大学 | Multicomponent self-organizing soft connecting assembly computer as well as self-organizing method thereof |
| CN1713196B (en) * | 2005-07-14 | 2011-08-31 | 上海交通大学 | Product ordering system based on automatic design grid |
| CN104360980A (en) * | 2014-12-04 | 2015-02-18 | 浪潮电子信息产业股份有限公司 | Multi-plane separating type multi-segment computer architecture |
| CN104572295A (en) * | 2014-12-12 | 2015-04-29 | 北京应用物理与计算数学研究所 | Structured grid data management method matched with high-performance computer system structure |
| CN104731531A (en) * | 2015-03-24 | 2015-06-24 | 浪潮集团有限公司 | Server node architecture design method for separated type high-capacity memory |
| WO2016201704A1 (en) * | 2015-06-19 | 2016-12-22 | 华为技术有限公司 | Optical switch control method and device |
| CN106547621A (en) * | 2016-10-21 | 2017-03-29 | 黄东 | A kind of gridding resource Optimization Scheduling under the conditions of large scale |
| WO2022100364A1 (en) * | 2020-11-13 | 2022-05-19 | 华为技术有限公司 | Method, apparatus and computing device for configuring compute mode |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| DE4117693A1 (en) * | 1991-05-29 | 1992-12-03 | Parsytec Computer Gmbh | Fault tolerant computer system - has function modules with redundant processing and communication built into three=dimensional communication system |
| US6275975B1 (en) * | 1997-01-16 | 2001-08-14 | Advanced Micro Devices, Inc. | Scalable mesh architecture with reconfigurable paths for an on-chip data transfer network incorporating a network configuration manager |
-
2003
- 2003-06-13 CN CNB031423361A patent/CN100351829C/en not_active Expired - Fee Related
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN100428167C (en) * | 2004-03-19 | 2008-10-22 | 国际商业机器公司 | Computer system, server constituting the same, job execution control method thereof, and program |
| CN100440832C (en) * | 2004-08-20 | 2008-12-03 | 清华大学 | Method for building self-organized network skeleton structure |
| CN100391149C (en) * | 2005-05-12 | 2008-05-28 | 中国科学院计算技术研究所 | Application managing sub-system in multi-node intelligent network application service system and method |
| CN1713196B (en) * | 2005-07-14 | 2011-08-31 | 上海交通大学 | Product ordering system based on automatic design grid |
| CN100435159C (en) * | 2006-08-24 | 2008-11-19 | 上海交通大学 | Method for implementing mesh based optical mesh emulation platform |
| CN101226518B (en) * | 2008-01-25 | 2010-11-03 | 华南理工大学 | Multicomponent self-organizing soft connecting assembly computer as well as self-organizing method thereof |
| CN104360980B (en) * | 2014-12-04 | 2017-08-11 | 浪潮电子信息产业股份有限公司 | A kind of many planar separation formula multi partition Computer Architectures |
| CN104360980A (en) * | 2014-12-04 | 2015-02-18 | 浪潮电子信息产业股份有限公司 | Multi-plane separating type multi-segment computer architecture |
| CN104572295A (en) * | 2014-12-12 | 2015-04-29 | 北京应用物理与计算数学研究所 | Structured grid data management method matched with high-performance computer system structure |
| CN104572295B (en) * | 2014-12-12 | 2016-06-01 | 北京应用物理与计算数学研究所 | It is matched with the structured grid data management process of high-performance calculation machine architecture |
| CN104731531A (en) * | 2015-03-24 | 2015-06-24 | 浪潮集团有限公司 | Server node architecture design method for separated type high-capacity memory |
| CN104731531B (en) * | 2015-03-24 | 2018-01-02 | 浪潮集团有限公司 | A kind of server node architecture design method of separate type high power capacity internal memory |
| WO2016201704A1 (en) * | 2015-06-19 | 2016-12-22 | 华为技术有限公司 | Optical switch control method and device |
| US10439736B2 (en) | 2015-06-19 | 2019-10-08 | Huawei Technologies Co., Ltd. | Optical switch control method, and apparatus |
| CN106547621A (en) * | 2016-10-21 | 2017-03-29 | 黄东 | A kind of gridding resource Optimization Scheduling under the conditions of large scale |
| WO2022100364A1 (en) * | 2020-11-13 | 2022-05-19 | 华为技术有限公司 | Method, apparatus and computing device for configuring compute mode |
Also Published As
| Publication number | Publication date |
|---|---|
| CN100351829C (en) | 2007-11-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11436400B2 (en) | Optimization method for graph processing based on heterogeneous FPGA data streams | |
| US10728091B2 (en) | Topology-aware provisioning of hardware accelerator resources in a distributed environment | |
| US8856794B2 (en) | Multicore runtime management using process affinity graphs | |
| CN100351829C (en) | Self-organizing dynamic network computer system structure | |
| CN102929725B (en) | Dynamic reconfiguration method of signal processing parallel computing software | |
| CN111210879B (en) | Hierarchical storage optimization method for super-large-scale drug data | |
| Zhang et al. | MrHeter: improving MapReduce performance in heterogeneous environments | |
| JP2023025146A (en) | Distributed type training method based on end-to-end self-adaptation, device, and apparatus | |
| CN102937964B (en) | Intelligent data service method based on distributed system | |
| Arfat et al. | Big data for smart infrastructure design: Opportunities and challenges | |
| Cong et al. | CPU-FPGA coscheduling for big data applications | |
| Muhammad Faseeh Qureshi et al. | RDP: A storage-tier-aware Robust Data Placement strategy for Hadoop in a Cloud-based Heterogeneous Environment | |
| WO2020040928A1 (en) | System for optimizing storage replication in a distributed data analysis system using historical data access patterns | |
| CN112379985A (en) | Computing task allocation method and device in cloud edge computing environment | |
| Song et al. | Building spatiotemporal cloud platform for supporting GIS application | |
| CN101290592B (en) | Realization method for multiple program sharing SPM on MPSOC | |
| CN116302574B (en) | Concurrent processing method based on MapReduce | |
| US20180246806A1 (en) | Memory management in multi-processor environments | |
| Zhao et al. | Gpu-accelerated cloud computing for data-intensive applications | |
| CN110415162B (en) | Adaptive graph partitioning method facing heterogeneous fusion processor in big data | |
| Fang et al. | A real-time and reliable dynamic migration model for concurrent taskflow in a GPU cluster | |
| Bang et al. | Accelerating I/O performance of ZFS-based Lustre file system in HPC environment | |
| JP2022531353A (en) | Equipment and methods for dynamically optimizing parallel computing | |
| Song et al. | Memory management optimization strategy in Spark framework based on less contention | |
| Liao et al. | Optimizing data query performance of Bi-cluster for large-scale scientific data in supercomputers |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20071128 Termination date: 20200613 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |