CN108156064A - Computer cluster system - Google Patents

Computer cluster system Download PDF

Info

Publication number
CN108156064A
CN108156064A CN201611110011.0A CN201611110011A CN108156064A CN 108156064 A CN108156064 A CN 108156064A CN 201611110011 A CN201611110011 A CN 201611110011A CN 108156064 A CN108156064 A CN 108156064A
Authority
CN
China
Prior art keywords
node
computer
data
host
ring topology
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
Application number
CN201611110011.0A
Other languages
Chinese (zh)
Inventor
庄克竣
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Mitac Computer Shunde Ltd
Shencloud Technology Co Ltd
Original Assignee
Mitac Computer Shunde Ltd
Shencloud Technology Co Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Mitac Computer Shunde Ltd, Shencloud Technology Co Ltd filed Critical Mitac Computer Shunde Ltd
Priority to CN201611110011.0A priority Critical patent/CN108156064A/en
Publication of CN108156064A publication Critical patent/CN108156064A/en
Pending legal-status Critical Current

Links

Classifications

    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L12/00Data switching networks
    • H04L12/28Data switching networks characterised by path configuration, e.g. LAN [Local Area Networks] or WAN [Wide Area Networks]
    • H04L12/42Loop networks
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L12/00Data switching networks
    • H04L12/28Data switching networks characterised by path configuration, e.g. LAN [Local Area Networks] or WAN [Wide Area Networks]
    • H04L12/42Loop networks
    • H04L2012/421Interconnected ring systems

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Small-Scale Networks (AREA)

Abstract

A kind of computer cluster system includes multiple computer nodes.Such computer node is divided into multiple groups, and each group includes multiple computer nodes that one first ring topology is collectively formed.The first host node that the one of which of such computer node of every one first ring topology corresponds to the first ring topology belonging to itself for one, and one second ring topology is collectively formed in such first host node, the one of which of such first host node is second host node for corresponding to second ring topology.The present invention can be not required to an additional computer node to manage whole system, and be compared with existing ring topology, have higher data transmission efficiency, can more meet the data transmission demand of real-time.

Description

Computer cluster system
Technical field
The invention relates to a kind of computer cluster systems, particularly relate to a kind of computer for forming multilayer ring topology Cluster system.
Background technology
Computer cluster system (computer cluster system) contains multiple computer node (computer Node), wherein the connection mode of such computer node i.e. the network topology of computer cluster system.
In general, the framework of network topology can substantially divide into centre management framework and two class of non-stop layer management framework. In centre management framework, star topology as shown in Figure 1 (star topology) is common are, wherein by a centralized computer Node 8 is directly connected to other multiple computer nodes 9, and is responsible for the management and monitoring of whole system.Above-mentioned centre management framework Although there is higher data transmission efficiency, whole system is managed because needing an additional centralized computer node 8, so hardware Cost is also higher.In non-stop layer management framework, ring topology as shown in Figure 2 (ring topology) is common are, wherein One is selected by such computer node 9 to be responsible for and monitor whole system.In this framework, because being not required to an additional centralized computer Node manages whole system, so hardware cost is relatively low;But the maximum data transmission path length between two computer nodes 9 is The half of 9 quantity of computer node, so data transmission is less efficient, when especially 9 quantity of computer node is more.On in view of It states, the network topology architecture of existing non-stop layer management also has improved space.
Invention content
Therefore, the purpose of the present invention is providing a kind of computer of the non-stop layer management with high data transmission efficiency Cluster system.
Then, computer cluster system of the present invention includes multiple computer nodes.Such computer node is divided into multiple groups Group, each group include multiple computer nodes that one first ring topology is collectively formed.
The one of which of such computer node of every one first ring topology corresponds to the first ring topology belonging to itself for one The first host node, and one second ring topology, the one of which of such first host node is collectively formed in such first host node The second host node for corresponding to second ring topology for one.
In some embodiments, each computer node has a node data, and in every one first ring topology, this first Host node penetrates two computer nodes other computer festivals received from body belonging to first ring topology adjacent with first host node The node data of point, and second host node receives through two first host nodes adjacent with second host node and is different from itself The node data of the computer node of other the first ring topologies of affiliated first ring topology.
In some embodiments, in every one first ring topology, each computer node regularly sends a data requirement To each adjacent computer node, each computer node returns after the data requirement from an adjacent computer node is received All node data itself stored are passed to the adjacent computer node, in second ring topology, every one first main section Point regularly sends a data requirement to each the first adjacent host node, every one first host node receive it is adjacent from one The first host node data requirement after, return all node data stored by itself immediately to the first adjacent main section Point.
In some embodiments, when second host node is each from two adjacent first host nodes in a data receives bout From the node data for receiving same computer node when, second host node is by judging those corresponding to same computer node The path-length relative to second host node of each node data is one relatively new in those node data to judge Node data and an older node data wherein should be less than the older node data compared with the path-length of new node data Path-length, and the path-length relative to second host node of each node data for the node data at this The summation of path-length and the path-length in second ring topology in first ring topology.
In some embodiments, the node data of each computer node includes the transmission being somebody's turn to do relative to second host node Path length, and the path-length is during the node data is sent to second host node, to be transmitted/connect The number of nodes of receipts.
In some embodiments, each node data includes a timestamp, when second host node is received back in a data When being respectively received the node data of same computer node from two the first adjacent host nodes in conjunction, second host node is according to this Those timestamps in a little node data come judge in those node data one compared with new node data and an older node money Material.
In some embodiments, which includes a monitoring information of the computer node.
Beneficial effects of the present invention are:It is compared with existing ring topology, the present invention has higher data transmission efficiency.
【Description of the drawings】
Fig. 1 is a block diagram, illustrates the computer cluster system of existing star topology;
Fig. 2 is a block diagram, illustrates the computer cluster system of existing ring topology;And
Fig. 3 is a block diagram, illustrates the computer cluster system of multilayer ring topology of the present invention.
【Specific embodiment】
Before the present invention is described in detail, it shall be noted that in the following description content, similar element is with identical volume It number represents.
Refering to Fig. 3, Fig. 3 illustrates the network topology architecture of an embodiment of computer cluster system of the present invention.The computer Cluster system includes 64 computer nodes 9, which is divided into 16 groups, each group Comprising four computer nodes 9, and one first ring topology 1 is formed, therefore share 16 the first ring topologies 1.Every one first The first main section that the one of which of four computer nodes 9 of ring topology 1 corresponds to the first ring topology 1 belonging to itself for one Point, and one second ring topology 2 is collectively formed in every four the first host nodes, therefore there are four the second ring topologies 2 altogether.Every 1 The second master that the one of which of four first host nodes of two ring topologies 2 corresponds to the second ring topology 2 belonging to itself for one Node;And a third ring topology 3 is collectively formed in the second all host nodes, and the one of which of such second host node is one The third host node of the corresponding third ring topology 3, and it is responsible for and monitors whole system by the third host node.
The network topology architecture of the present embodiment can be considered one three layers of ring topology framework, wherein first layer correspond to this ten Six the first ring topologies 1, the second layer correspond to this four the second ring topologies 2, and third layer corresponds to the third ring topology 3.
Illustrate that the third host node collects the mode of the node data of each computer node 9 below, wherein each computer festival The node data of point 9 includes the monitoring information of the computer node 9, including cpu temperature, memory temperature etc..
In every one first ring topology 1, each computer node 9 regularly sends a data requirement to each adjacent electricity Encephalomere point 9;Each computer node 9 returns itself mesh immediately after the data requirement from an adjacent computer node 9 is received Preceding stored all node data are any adjacent with penetrating including the node data of itself to the adjacent computer node 9 The node data of other computer nodes 9 received by computer node 9;In this way, every one first host node can store itself institute Belong to the node data of each computer node 9 of the first ring topology 1.
Similarly, in every one second ring topology 2, every one first host node regularly sends a data requirement to each The first adjacent host node;Every one first host node after the data requirement from a first adjacent host node is received, with Itself all node data stored at present is returned to the first adjacent host node, it is cyclic annular including first belonging to itself The node data of any computer node 9 of topology 1, with penetrating other computer festivals received by any the first adjacent host node The node data of point 9;In this way, every one second host node can store that the second ring topology 2 belonging to itself connected four the The node data of each computer node 9 of one ring topology 1.
Similarly, in the third ring topology 3, every one second host node regularly sends a data requirement to each phase The second adjacent host node;Every one second host node is after the data requirement from a second adjacent host node is received, immediately Itself all node data stored at present is returned to the second adjacent host node, is opened up including the second ring-type belonging to itself The node data of any computer node 9 of 2 any first ring topologies 1 connected is flutterred, with penetrating any the second adjacent master The node data of other computer nodes 9 received by node;In this way, the third host node can receive 64 computers The node data of node 9.
In every one first ring topology 1, each node data can be made to include a path-length, wherein, Mei Yijie The path-length of point data for the node data in transmission process by the number of nodes of transmission/reception, when the first main section O'clock when being respectively received the node data of same computer node 9 from 2 adjacent computer festival points 9 during a data receives bout, this One host node is by judging that those are long corresponding to the corresponding transmission path of each node data institute of same computer node 9 Degree, that is, those corresponding to same computer node 9 node data be sent to first host node during, passed The number of nodes for sending/receiving is transferred into the biography of first host node with those corresponding to the node data of same computer node Defeated path length judges that first host node is respectively received from 2 adjacent computer festival points 9 corresponding to same computer node 9 Node data in one compared with new node data and an older node data, and abandon the older node data, wherein this is newer The path-length of node data is less than the path-length of the older node data.For example, refering to Fig. 3, computer The node data of node 9b can be by the path transmission of " computer node 9b → first host node 9c " to the first host node 9c, i.e. path Length is the length of node or the road of " computer node 9b → computer node 9a → computer node 9p → first host node 9c " Diameter is transmitted to the first host node 9c, i.e. length of the path length for three nodes.If the first host node 9c is received in a data Respectively since above-mentioned two path receives the node data of computer node 9b in bout, then with from path " computer node 9b → the first Node data received by host node 9c " is is somebody's turn to do compared with new node data.
Similarly, in every one second ring topology 2, when second host node is adjacent from two in a data receives bout When first host node is respectively received the node data of corresponding same computer node 9, second host node is by judging that those are right Each node data of same computer node 9 is answered to be respectively transmitted to the path-length of second host node to judge this Being somebody's turn to do compared with new node data and the older section in the node data corresponding to same computer node 9 received by the second host node Point data, and the older node data is abandoned, the older node money should be wherein less than compared with the path-length of new node data The path-length of material, and the path-length relative to second host node of each node data is the node data The summation of path-length and the path-length in second ring topology 2 in first ring topology 1.
Similarly, in the third ring topology 3, when the third host node is adjacent from two in a data receives bout When second host node is respectively received the node data of same computer node 9, the third host node is by judging that those are corresponded to Each node data of same computer node 9 be respectively transmitted to the path-length of the third host node judge this Three host nodes received those corresponding in the node data of same computer node 9 should compared with new node data and the older section Point data, and the older node data is abandoned, the older node money should be wherein less than compared with the path-length of new node data The path-length of material, wherein, the path-length of each node data transmission to the third host node is provided for the node Expect the path-length in first ring topology 1, the path-length in second ring topology 2 and at this The summation of path-length in third ring topology 3.
In addition, in another embodiment, each node data can also be made to include a timestamp, in this way, as third master Node is respectively received the node data of same computer node 9 in a data receives bout from two the second adjacent host nodes When, the third host node can according to such timestamp in the two nodes data come judge in the two nodes data this is newer Node data and the older node data.
Particularly, it is compared with existing ring topology, the ring topology framework of multilayer shortens two computer in network topology The maximum data transmission path length of node.Specifically, the existing ring-type formed by 64 computer nodes is opened up For flutterring, the maximum data transmission path length of two computer nodes is 32.And to the shellring shape topology frame of the present embodiment For structure, the maximum data transmission path length of two computer nodes is ten.For example, refering to Fig. 3, data transmission path length Path for ten has:" computer node 9a → computer node 9b → computer node 9c → computer node 9d → computer node 9e → electricity Encephalomere point 9f → computer node 9g → computer node 9h → computer node 9i → computer node 9j → computer node 9k ".In addition, appoint One computer node 9 and the maximum data transmission path length of management, the third host node of monitoring whole system are six, such as road Diameter " computer node 9a → computer node 9b → computer node 9c → computer node 9d → computer node 9e → computer node 9f → electricity Encephalomere point 9g ".
Although illustrate the multilayer ring topology of computer cluster system of the present invention in the present embodiment with shellring shape topology Framework, but the tool usually intellectual of this case technical field should can be understood the layer of the multilayer ring topology by above description Number can be not limited to the numerical value of three or two, four or bigger.If in addition, the computer node of the computer cluster system is total Quantity is NK, and wherein N is the quantity of the computer node of single a ring topology in each layer, and K is the number of plies, then two computer festivals The length in the maximum data transmission path of point is " N ÷ 2+N × (K-1) ", and any computer node is with managing, the entire system of monitoring The maximum data transmission path length of the computer node of system is " 2 × K of N ÷ ".
The multilayer ring-type topological structure of computer cluster system of the present invention can be applied to a frame type computer cluster system (rack server cluster system), wherein each computer node includes a baseboard management controller (baseboard Management controller, BMC), which is used to monitor the functioning condition of the computer node, such as Monitor the current cpu temperature of the computer node, memory temperature etc..The frame type computer cluster system is provided with multiple outer Portion's fan, and need a to control comprising the node data of the data such as cpu temperature, memory temperature according to each computer node The rotating speed of other fan.
In existing ring topology framework, the maximum of computer node of any computer node with being responsible for monitoring whole system Data transmission path is the half of computer number of nodes, so when computer number of nodes is excessive, is responsible for the entire system of monitoring The computer node of system, and can not root in real time easily because need to be expended to collect the temperature data of all computer nodes the excessive time The rotating speed of such fan is controlled according to the temperature data of all computer nodes.
It is compared with existing ring topology, in the multilayer ring-type topological structure of the present invention, any computer node is with being responsible for The maximum data transmission path length for monitoring the computer node of whole system is relatively small, therefore is responsible for monitoring whole system Computer festival point collects the time relatively fewer many expended needed for the temperature data of all computer nodes.With 64 computer festivals For point, the maximum data transmission path length of existing ring topology is 32, and the maximum of shellring shape topology Data transmission path length is six, so efficiency of the shellring shape topology on the temperature data for collecting all computer nodes is apparent Higher than existing ring topology.Therefore, it is compared with existing ring topology, multilayer ring topology of the invention can more be reached in real time Ground is according to the temperature data of all computer nodes come the effect of controlling the rotating speed of such fan.
In conclusion computer cluster system of the present invention, it, can not by all computer nodes of multilayer ring-type Topology connection An additional computer node is needed to manage whole system, and is compared with existing ring topology, there is higher data transmission efficiency, The data transmission demand of real-time can more be met, therefore the purpose of the present invention can be reached really.
Above in conjunction with attached drawing to the present invention specific embodiment and embodiment elaborate, but cannot with restriction The scope of the present invention, the impartial modification and transformation made in scope of the present invention patent, should all belong to patent of the present invention In the range of.

Claims (7)

1. a kind of computer cluster system, it is characterised in that:Comprising:
Multiple computer nodes, are divided into multiple groups, and each group includes multiple electricity that one first ring topology is collectively formed Encephalomere point, the one of which of such computer node of every one first ring topology correspond to the first ring topology belonging to itself for one First host node, and one second ring topology is collectively formed in such first host node, the one of which of such first host node is One corresponds to the second host node of second ring topology.
2. computer cluster system according to claim 1, it is characterised in that:Each computer node is provided with a node Material, in every one first ring topology, first host node through the two computer nodes adjacent with first host node received from The node data of other computer nodes of the first ring topology belonging to body, and second host node penetrates and the second host node phase Two first adjacent host nodes, which receive, is different from the computer node of other the first ring topologies of the first ring topology belonging to itself Node data.
3. computer cluster system according to claim 2, it is characterised in that:It is each in every one first ring topology Computer node regularly sends a data requirement to each adjacent computer node, and each computer node is being received from a phase After the data requirement of adjacent computer node, all node data itself stored are returned to the adjacent computer node, at this In second ring topology, every one first host node regularly sends a data requirement to each the first adjacent host node, each First host node returns all stored by itself immediately after the data requirement from a first adjacent host node is received Node data gives the first adjacent host node.
4. computer cluster system according to claim 2, it is characterised in that:When second host node is received in a data When being respectively received the node data of same computer node from two the first adjacent host nodes in bout, which passes through Judge that those come corresponding to the path-length relative to second host node of each node data of same computer node Judge that one in those node data, wherein should be compared with the transmission of new node data compared with new node data and an older node data Path length is less than the path-length of the older node data, and each node data relative to second host node Path-length is for path-length of the node data in first ring topology and in second ring topology Path-length summation.
5. computer cluster system according to claim 4, it is characterised in that:The node data packet of each computer node Path-length containing this relative to second host node, and the path-length for the node data be sent to this During two host nodes, the number of nodes that is transmitted/receives.
6. computer cluster system according to claim 2, it is characterised in that:Each node data includes a timestamp, When second host node is respectively received same computer node in a data receives bout from two the first adjacent host nodes During node data, those timestamps of second host node in those node data are judged in those node data One compared with new node data and an older node data.
7. computer cluster system according to claim 2, it is characterised in that:The node data includes the computer node One monitoring information.
CN201611110011.0A 2016-12-06 2016-12-06 Computer cluster system Pending CN108156064A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201611110011.0A CN108156064A (en) 2016-12-06 2016-12-06 Computer cluster system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201611110011.0A CN108156064A (en) 2016-12-06 2016-12-06 Computer cluster system

Publications (1)

Publication Number Publication Date
CN108156064A true CN108156064A (en) 2018-06-12

Family

ID=62467931

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201611110011.0A Pending CN108156064A (en) 2016-12-06 2016-12-06 Computer cluster system

Country Status (1)

Country Link
CN (1) CN108156064A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110535706A (en) * 2019-09-02 2019-12-03 成都吉胜科技有限责任公司 A kind of automatic coordinated dispatching method based on local area network
CN113268337A (en) * 2021-07-20 2021-08-17 杭州朗澈科技有限公司 Method and system for Pod scheduling in Kubernetes cluster

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110535706A (en) * 2019-09-02 2019-12-03 成都吉胜科技有限责任公司 A kind of automatic coordinated dispatching method based on local area network
CN113268337A (en) * 2021-07-20 2021-08-17 杭州朗澈科技有限公司 Method and system for Pod scheduling in Kubernetes cluster

Similar Documents

Publication Publication Date Title
CN106126346B (en) A kind of large-scale distributed data collection system and method
CN100452794C (en) Master node selection in clustered node configurations
CN103236949B (en) Monitoring method, device and the system of a kind of server cluster
CN106663030B (en) Scalable failover communication in distributed clusters
CN105337780B (en) A kind of server node configuration method and physical node
US8880931B2 (en) Method, distributed system and computer program for failure recovery
CN106375462A (en) Method and device for realizing message persistence in distributed message system
Kutten et al. Sublinear bounds for randomized leader election
Ramanathan et al. Delivery of time-critical messages using a multiple copy approach
CN102025550A (en) System and method for managing data in distributed cluster
KR102508817B1 (en) High availability distribution intelligence system using message transmission bus
CN109845192B (en) Computer system and method for dynamically adapting a network and computer readable medium
CN106878382A (en) Dynamically change the method and device of cluster scale in a kind of distributed arbitration program cluster
CN112788798A (en) Master-slave negotiation multi-gateway intelligent Zigbee network system
CN106953926A (en) A kind of method for routing and device
CN109639773A (en) A kind of the distributed data cluster control system and its method of dynamic construction
Kutten et al. Sublinear bounds for randomized leader election
CN109728941A (en) A kind of block chain leader election method and device thereof
CN108156064A (en) Computer cluster system
TWI633773B (en) Computer cluster system
US20050022048A1 (en) Fault tolerance in networks
CN108809680A (en) A kind of method and apparatus of equipment management
CN103677994B (en) Distributed data processing system, device and method
CN104270466A (en) Data reporting method and related equipment
Kanrar et al. A new voting-based mutual exclusion algorithm for distributed systems

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: 20180612

RJ01 Rejection of invention patent application after publication