WO2001002972A1 - A scalable computer system - Google Patents
A scalable computer system Download PDFInfo
- Publication number
- WO2001002972A1 WO2001002972A1 PCT/AU2000/000796 AU0000796W WO0102972A1 WO 2001002972 A1 WO2001002972 A1 WO 2001002972A1 AU 0000796 W AU0000796 W AU 0000796W WO 0102972 A1 WO0102972 A1 WO 0102972A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- nodes
- network
- small
- world
- neighborhoods
- Prior art date
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F15/00—Digital computers in general; Data processing equipment in general
- G06F15/16—Combinations of two or more digital computers each having at least an arithmetic unit, a program unit and a register, e.g. for a simultaneous processing of several programs
- G06F15/163—Interprocessor communication
- G06F15/173—Interprocessor communication using an interconnection network, e.g. matrix, shuffle, pyramid, star, snowflake
- G06F15/17356—Indirect interconnection networks
- G06F15/17368—Indirect interconnection networks non hierarchical topologies
- G06F15/17375—One dimensional, e.g. linear array, ring
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L45/00—Routing or path finding of packets in data switching networks
- H04L45/02—Topology update or discovery
- H04L45/04—Interdomain routing, e.g. hierarchical routing
Definitions
- This invention relates to computer systems and more particularly, but not exclusively, to a computer system which is required to have a large number of computer processors for use in large scale system applications.
- the present invention has application to scalable telecommunications and intelligent networks.
- the Internet is a large computer system, but it is a hierarchical system and hence does not address the issue of massive scalability.
- Other computer system architectures that are known are only scalable in many orders of magnitude less than what will be required in the near future in telecommunications and intelligent networks.
- the computer system 10 includes hardware 12, an operating system 14, a display 16 and a keyboard 18 and a suite of programs 20 which include application programs 22 (eg. in programming languages, Erlang and C), sourced programs 24, run-time programs 26, a library 28, and a database 30.
- application programs 22 eg. in programming languages, Erlang and C
- sourced programs e.g. in programming languages, Erlang and C
- run-time programs e.g. in programming languages, Erlang and C
- the system may be linked to an external database 32 if required.
- the single node provides reasonable system development facilities, including an Erlang real time environment, or interpretive environment. However, this is achieved at the expense of potential performance owing to the interpreter/operating system layers.
- the single node computer system of Figure 1 may be linked to other similar nodes by an asynchronous transfer node (ATM) switch, such as the AXD-301 switch with satisfactory performance.
- ATM asynchronous transfer node
- this switch has a scalability of 1 :30, which is orders of magnitude less than that which is required for ultra high communication bandwidths.
- FIG. 2 there is shown a split node computer system in which an OTP node 34 is split into two closely couple nodes: a COTS (commodity of the shelf) system 40 and a multi-processor (MP) Erlang Engine 50.
- the COTS system is essentially the base system and may comprise of a UNIX operating platform 41 , application programs (eg. in C, C++, Java and Erlang) 42, a disc drive 43, graphics 44, an Internet modem 45, an Internet interface (TCP/IP) 46 and an input/output interface (I/O) 47 for communicating with the Erlang Engine 50.
- application programs eg. in C, C++, Java and Erlang
- TCP/IP Internet interface
- I/O input/output interface
- the Erlang Engine 50 is a shared memory MP system running software 52 and 54 in Erlang on top of an optimized message passing kernel (56) such as QNX.
- the Erlang Engine 50 also has an I/O interface 58 for communicating with the COTS system 40.
- One processor of the MP set can be devoted to monitoring software, the remainder to functional processing.
- the split node OTP system 34 of Figure 2 may form part of a network that includes a plurality of regional processors (RP) 61 and support processors (SP) 62 for operators.
- the regional and support processors 61 , 62 are connected to central processors CP A 63 and CP B 64 and to each other by a high speed RP bus 65.
- the MP Erlang Engine 50 includes a high speed interface 59 for communicating with the central processors CP A 63 and CP B 64.
- the split node OTP system 34 can be linked to other computer systems by a switch 70 such as an AXE-10.
- an AXE programming system (APS) 72 may be provided.
- the interface shown to the AXE-10 may be implemented as a high speed Ethernet, primarily due to the availability of the Ethernet PLEX (Programming Language for Exchanges) blocks existing on AXE-10.
- the Erlang to PLEX interface has been demonstrated in two modes, firstly with the AXE-10 controlling the links, as in a call forwarding application. The second mode is with OTP in control, with an application such as remote changes to AXE-10 tariff tables.
- the scalability limit of the Erlang Engine is a maximum of eight processors, given that it is a shared memory environment.
- the present invention provides a telecommunications system having a plurality of nodes, such that the plurality of nodes are clustered in a plurality of interconnected neighborhoods, the system characterised in that a relatively small number of cross-links are provided between nodes of different neighborhoods.
- the present invention provides a large scale computer system including a multiplicity of nodes, each node having a plurality of interconnected processors, said nodes being arranged in a network with neighbouring sets of nodes of the network forming neighbourhoods of fully interconnected nodes, wherein random links are provided between nodes of different neighbourhoods in the network whereby each processor of the system can communicate effectively with other processors regardless of their location in the network without full connectivity in the network.
- the present invention has come about through the realisation that problems associated with telecommunication and intelligent network scalability and/or computer architecture design, can be addressed by applying the "small world" principle.
- the "small world” concept was derived from the "six degrees of separation” principle.
- Six degrees of separation is a theory that a person may reach anyone in the world through at the most six personal contacts. This theory was explored through an Internet site based upon the actor Kevin Bacon. This site was created out of curiosity, but was a very useful sociological experiment for illustrating the six degrees of separation principle. The site however, gave no hint as to any underlying structure of the people involved in the interconnections.
- the key to the underlying structure may be derived from studies into epidemics. For example diseases, such as the Black Death spread with great rapidity from one part of the world to another: from the Middle East to Europe. In that situation, the epidemic mechanism was that of an area that already had the disease within a defined group, and a single vector, such as ship-borne rats travelling to another defined group.
- the present invention utilizes a "small-world" network architecture, in which a relatively small number of random cross-links of nodes or vertices in a network can result in small characteristic path lengths, for the transfer of messages between nodes or vertices in a telecommunications/computer network regardless of their location.
- FIG. 1 is a diagram of a prior art single node computer system, which has limited scalability
- Figure 2 is a diagram of a prior art split node computer system
- Figure 3 is a diagram showing the difference between a "small world” network and regular ring and random networks
- Figure 4 is a graph showing variations in clustering and average path length with increasing randomness in a ring network.
- Figure 5 is a diagram of a four-node computer system having four processors per node.
- Figure 6 is a diagram of a small-world architecture for a computer system in accordance with the invention.
- the "small world” principle is usually considered to apply to many biological and social networks, as these systems generally exhibit properties that are not completely regular or completely random but somewhere in between.
- infectious diseases exhibit "small world” properties.
- the present invention applies this small world principle to telecommunications/computer networks.
- a comparison of the performance of a small world telecommunications network, to that of regular and random telecommunications networks will now be provided. Examples of regular, small- world and random networks are shown in Figure 3.
- a small world telecommunications network in accordance with the present invention is one in which vertices or nodes in local “neighborhoods” or “clusters" are interconnected to each other, and a relatively small number of random links or connections are provided between nodes of different neighborhoods of the network.
- L(p) measures the typical separation between two vertices in a network (a global property) and C(p) measures the clustering or connections in a typical neighborhood (a local property), where p is a probability factor.
- Such a logarithmic curve approximates the curve L(p)/L(0) in Figure 4.
- the number of nodes in the network, the number of nodes in the neighborhood, the number of nodes per neighborhood and the number of cross-links in the network may each be varied for different applications, provided that the network functions as a small-world network with large clustering C(p) and small average path lengths L(p).
- the cross-links in the small-world network of the present invention may be chosen completely at random.
- a pseudo-random selection process may be used to select the cross-links between neighborhoods to convert a regular network into a highly clustered small-world network with a relatively small average path length.
- an acceptable result in terms of the number of hops per inter-process message, from a system process point-of-view is under two.
- a mean connectivity falling substantially in the range from about 1.5 to about 2.0, and preferably about 1.6, may be achieved by appropriate choice of the number of nodes per neighborhood, the connectivity of each neighborhood and/or the number of cross-links relative to the total number of nodes in the network.
- the number of interconnected processors in each node of the computer system may also vary for different applications. With the development of photonics, it is envisaged that up to about 256 nodes, each containing up to eight processors per node may be fully connected together by optical fibres and a high speed switch to form a single neighborhood in a small-world network. With a large number of such neighborhoods in the small-world network connected by a relatively small number of cross-links, it is possible to achieve a massively scalable computer system with a very large number of total processors in the system which are able to communicate with each other in an effective manner.
- Figure 6 shows a small-world telecommunications network in which there are 20 nodes arranged in a ring layout.
- the 20 nodes of the ring network can be considered to form five neighborhoods of nodes A, B, C, D and E in which there are four nodes (A, to A 4 ; B 1 to B 4 ; C, to C 4 ; D, to D 4 ; and E to E 4 ) per neighborhood.
- Each node eg A 2 is connected to adjacent nodes A v A 3 by an edge link EL and alternate nodes eg A,, A 3 ; A 2 , A 4 ; A 3 , B, are connected to each other by a small loop link LL.
- a network connected only in this manner would normally be termed a regular network in which messages from one node, eg A,, must pass through a relatively large number of links to reach a node at the opposite side of the network, eg C 4 .
- a small-world network such as shown in Figure 6 differs from a regular network in that a relatively small number of random cross-links are provided between neighborhoods of nodes.
- Cross-link CL directly connects nodes A 2 and C, with each other;
- cross-link CL 2 directly connects nodes B 2 and E 3 with each other and
- cross-link CL 3 directly connects nodes D, and E 4 with each other. It will be apparent from Figure 6 that there is a marked increase in connectivity with only a relatively small number of cross-links between neighborhoods of nodes.
- a message to be sent from node A, to C 2 can pass along an edge link EL to node A 2 , along cross-link CL, to C, and then along an edge link to C 2 .
- This is a clear improvement over a regularly connected network, where the message would pass along four loop links to C, and an edge link EL to C 2 .
- a message from B 4 to E 2 can pass along a loop link to B 2 , along cross-link CL 2 to E 4 and then along another loop link LL to E 2 , rather than along five loop links.
- a small-world network is an architecture that can be used to link together a relatively large number of computer processors while retaining effective connectivity between the processors.
- each neighborhood of Figure 6 corresponds to the four-node/four-process per node computer system shown in Figure 5.
- Figure 5 there is shown a system that demonstrates it is possible to link up several MP Erlang Engines of the type shown in Figure 2 through a high-speed blocking switch.
- Each node of the computer system of Figure 5 is an Erlang Engine 150 with four processors 152 running on a QNX kernel 156.
- Each Erlang Engine 150 may include a link to an OTP system 134 in similar manner to split node system of Figure 2 and may also have a local Internet connection 145.
- the Erlang Engines 150 are connected to each other by optical interfaces 160, optical fibres 165 and a common high speed non-blocking optical switch 170.
- Each optical interface 160 may include at least one optical receiver 161 and one or more optical transmitters 162.
- interface blocks 160 can handle 32 separate wavelengths: this is predicted to rise to 256 with further developments in photonics. This will permit up to 256 nodes (Erlang Engines) in a computer system such as that of Figure 5, as each node can be essentially addressed by an individual wavelength.
- Erlang Engines Erlang Engines
- a key attribute of the system of Figure 5 is that the interface blocks can transmit one or more wavelengths, up to 256, with the same data simultaneously as a multicast facility. However, each node only receives on one wavelength: its address.
- each Erlang Engine 150 has its own dedicated optical fibre link 165 to the central switch 170.
- the small-world network of Figure 6 is, however, only a relatively small-scale example of such a network.
- at least 500 neighborhoods of nodes each having up to 256 fully connected Erlang Engines per node with 8 processors per Erlang Engine could be linked together in a small-world network to achieve a total system of over one million processors, with the small-world network architecture providing effective connectivity between the nodes with only a relatively small number of cross-links, say 50, between the neighborhoods.
- the total number of individual Erlang processors running on such a system will be of the order of 256,000,000 assuming 2000 active processors per Erlang Engine node.
- microbank applications whereby the present invention may be used by financial institutions in order to allow them to economically handle extremely large numbers of transactions, particularly relating to very small amounts.
- the size of the network, the neighborhoods in the network, the number of nodes per neighborhood, the number of cross-links between neighborhoods and the number of processors per node may be varied for different applications.
Abstract
Description
Claims
Priority Applications (4)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US10/030,342 US7752337B1 (en) | 1999-06-30 | 2000-06-30 | Scalable computer system |
JP2001508698A JP4708647B2 (en) | 1999-06-30 | 2000-06-30 | Extensible computer system |
AU55138/00A AU5513800A (en) | 1999-06-30 | 2000-06-30 | A scalable computer system |
EP00940055A EP1190333A1 (en) | 1999-06-30 | 2000-06-30 | A scalable computer system |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
AUPQ1286 | 1999-06-30 | ||
AUPQ1286A AUPQ128699A0 (en) | 1999-06-30 | 1999-06-30 | A scalable computer system |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2001002972A1 true WO2001002972A1 (en) | 2001-01-11 |
Family
ID=3815475
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/AU2000/000796 WO2001002972A1 (en) | 1999-06-30 | 2000-06-30 | A scalable computer system |
Country Status (4)
Country | Link |
---|---|
US (1) | US7752337B1 (en) |
JP (1) | JP4708647B2 (en) |
AU (1) | AUPQ128699A0 (en) |
WO (1) | WO2001002972A1 (en) |
Families Citing this family (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7953789B2 (en) * | 2002-09-06 | 2011-05-31 | International Business Machines Corporation | Who, what, where, when information supply facility |
JP4621424B2 (en) * | 2003-11-20 | 2011-01-26 | 独立行政法人科学技術振興機構 | Programmable logic circuit and wiring structure of programmable logic circuit |
JP2005270909A (en) * | 2004-03-26 | 2005-10-06 | Advanced Telecommunication Research Institute International | Hardware simulator |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP0404339A2 (en) * | 1989-06-22 | 1990-12-27 | Digital Equipment Corporation | Routing apparatus and method for high-speed mesh connected local area network |
EP0637152A1 (en) * | 1993-07-30 | 1995-02-01 | International Business Machines Corporation | Method and apparatus to speed up the path selection in a packet switching network |
US5602839A (en) * | 1995-11-09 | 1997-02-11 | International Business Machines Corporation | Adaptive and dynamic message routing system for multinode wormhole networks |
Family Cites Families (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5371852A (en) * | 1992-10-14 | 1994-12-06 | International Business Machines Corporation | Method and apparatus for making a cluster of computers appear as a single host on a network |
US5577204A (en) * | 1993-12-15 | 1996-11-19 | Convex Computer Corporation | Parallel processing computer system interconnections utilizing unidirectional communication links with separate request and response lines for direct communication or using a crossbar switching device |
FR2737030B1 (en) * | 1995-07-21 | 1997-08-14 | Bull Sa | METHOD FOR TRANSFERRING MESSAGES IN A MULTINODAL COMPUTER SYSTEM |
US6915271B1 (en) * | 1998-03-11 | 2005-07-05 | The Product Engine, Inc. | Method and system for delivering redeeming dynamically and adaptively characterized promotional incentives on a computer network |
US6675170B1 (en) * | 1999-08-11 | 2004-01-06 | Nec Laboratories America, Inc. | Method to efficiently partition large hyperlinked databases by hyperlink structure |
US7139262B1 (en) * | 2000-05-18 | 2006-11-21 | Bbn Technologies Corp. | Systems and methods for creating wireless small world networks |
GB2378279A (en) * | 2001-07-31 | 2003-02-05 | Hewlett Packard Co | Distribution of information |
WO2003034669A1 (en) * | 2001-10-17 | 2003-04-24 | British Telecommunications Public Limited Company | Network location management system |
-
1999
- 1999-06-30 AU AUPQ1286A patent/AUPQ128699A0/en not_active Abandoned
-
2000
- 2000-06-30 US US10/030,342 patent/US7752337B1/en not_active Expired - Fee Related
- 2000-06-30 WO PCT/AU2000/000796 patent/WO2001002972A1/en not_active Application Discontinuation
- 2000-06-30 JP JP2001508698A patent/JP4708647B2/en not_active Expired - Fee Related
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP0404339A2 (en) * | 1989-06-22 | 1990-12-27 | Digital Equipment Corporation | Routing apparatus and method for high-speed mesh connected local area network |
EP0637152A1 (en) * | 1993-07-30 | 1995-02-01 | International Business Machines Corporation | Method and apparatus to speed up the path selection in a packet switching network |
US5602839A (en) * | 1995-11-09 | 1997-02-11 | International Business Machines Corporation | Adaptive and dynamic message routing system for multinode wormhole networks |
Also Published As
Publication number | Publication date |
---|---|
JP4708647B2 (en) | 2011-06-22 |
JP2003504920A (en) | 2003-02-04 |
US7752337B1 (en) | 2010-07-06 |
AUPQ128699A0 (en) | 1999-07-22 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Yang et al. | Permutation capability of optical multistage interconnection networks | |
Mendia et al. | Optimal broadcasting on the star graph | |
Li et al. | Parallel computing using optical interconnections | |
Lea | Bipartite graph design principle for photonic switching systems | |
US7752337B1 (en) | Scalable computer system | |
EP1190333A1 (en) | A scalable computer system | |
AU5513800A (en) | A scalable computer system | |
Yang et al. | A new design for wide-sense nonblocking multicast switching networks | |
Suganuma et al. | Flexible network layer in dynamic networking architecture | |
Shao | On k-broadcasting in graphs | |
Al-Shammari et al. | Midimew Connected Torus Network for Next Generation Massively Parallel Computer System | |
Yang et al. | Routing permutations with link-disjoint and node-disjoint paths in a class of self-routable interconnects | |
Avoussoukpo et al. | Polyvalent wireless communication system (PWCS); a potentially useful technology for opportunistic networks | |
KR100288349B1 (en) | Compact Programmable Parallel Processing Module | |
Jaros | Evolutionary optimization of multistage interconnection networks performance | |
Haney et al. | Sliding-banyan network performance analysis | |
Giglmayr | d-dimensional (d≥ 3) shuffle interconnections | |
Giglmayr | Organization of k× k switches (k≥ 4) interconnected by d-dimensional (d≥ 2) regular optical patterns | |
Liang et al. | Permutation routing in all-optical product networks | |
Sahoo et al. | A novel hybrid data center architecture employing optically-switched WDM LANs over electrical fat trees | |
Matsuda | Theoretical limitations of a Hopfield network for crossbar switching | |
Araiijo et al. | Parallel training for neural networks using PVM with shared memory | |
Xu et al. | Optimal bandwidth utilization of all-optical ring with a converter of degree 4 | |
Ahamed et al. | Intelligent Networks/2 (IN/2) | |
Taka et al. | Design model of a twisted and folded Clos network with multi-step grouped intermediate switches guaranteeing admissible blocking probability |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
AK | Designated states |
Kind code of ref document: A1 Designated state(s): AE AG AL AM AT AU AZ BA BB BG BR BY BZ CA CH CN CR CU CZ DE DK DM DZ EE ES FI GB GD GE GH GM HR HU ID IL IN IS JP KE KG KP KR KZ LC LK LR LS LT LU LV MA MD MG MK MN MW MX MZ NO NZ PL PT RO RU SD SE SG SI SK SL TJ TM TR TT TZ UA UG US UZ VN YU ZA ZW |
|
AL | Designated countries for regional patents |
Kind code of ref document: A1 Designated state(s): GH GM KE LS MW MZ SD SL SZ TZ UG ZW AM AZ BY KG KZ MD RU TJ TM AT BE CH CY DE DK ES FI FR GB GR IE IT LU MC NL PT SE BF BJ CF CG CI CM GA GN GW ML MR NE SN TD TG |
|
121 | Ep: the epo has been informed by wipo that ep was designated in this application | ||
DFPE | Request for preliminary examination filed prior to expiration of 19th month from priority date (pct application filed before 20040101) | ||
WWE | Wipo information: entry into national phase |
Ref document number: 10030342 Country of ref document: US |
|
WWE | Wipo information: entry into national phase |
Ref document number: 2000940055 Country of ref document: EP |
|
WWE | Wipo information: entry into national phase |
Ref document number: 55138/00 Country of ref document: AU |
|
WWP | Wipo information: published in national office |
Ref document number: 2000940055 Country of ref document: EP |
|
REG | Reference to national code |
Ref country code: DE Ref legal event code: 8642 |
|
WWW | Wipo information: withdrawn in national office |
Ref document number: 2000940055 Country of ref document: EP |