WO2014067268A1

WO2014067268A1 - Node partition dividing method, device and server

Info

Publication number: WO2014067268A1
Application number: PCT/CN2013/074395
Authority: WO
Inventors: 廖义祥; 吴登奔; 张羽
Original assignee: 华为技术有限公司
Priority date: 2012-10-29
Filing date: 2013-04-19
Publication date: 2014-05-08
Also published as: CN102932175B; CN102932175A

Abstract

Disclosed are a node partition dividing method, device and server, which relate to the technical field of electronic information, and can automatically analyse and obtain a partition scheme with a minimum hop count of the QPI according to a topological structure of a system, so that the influence of subjective factors during manual partition division is avoided, and a reduction in operating rate caused by improper partition is alleviated, thereby alleviating the reduction in system performance.The method of the present invention comprises: acquiring a topological structure of a system and the number of nodes participating in partition, the system comprising at least three nodes, and each node comprising at least two CPUs;determining connection information according to the topological structure of the system, the connection information comprising a connection relationship between each CPU and other CPUs in the system;and determining a partition scheme according to the number of nodes participating in partition and the connection information, wherein the number of nodes in the partition scheme is the number of nodes participating in partition.The present invention is applicable to server partition.

Description

Method, device and server for dividing node partition

Technical field

The present invention relates to the field of electronic information technology, and in particular, to a method, an apparatus, and a server for dividing a node partition. Background technique

With the development of electronic information technology, the amount of data in various electronic information systems is increasing. For example: In order to support 3G networks, wireless communication systems must carry more data traffic; online storage systems in the Internet can support For many users, it is necessary to expand the data capacity of the system. The requirements for data carrying capacity and fault tolerance performance of various electronic information systems are getting higher and higher.

As a result, the server interconnection technology is derived, that is, the CPUs of the plurality of nodes are connected to each other through a technical means such as QP I (Quin-Calc Int erconnec t), so that a plurality of nodes that work independently can be combined. A whole (that is, a partition), and the combined entity as an executive body to undertake the work on the original single node, thereby improving the server data capacity and fault tolerance. For example: In Figure 1, node 0, node 1, node 2, node 3, these four nodes are divided into the same partition, and the CPUs of each node are connected by QPI, which is equal to the four nodes. A total of 8 CPUs make up a new server.

In the prior art, since the number of interfaces of the server and the CPU is limited, in actual applications, it is often impossible to connect each CPU to other CPUs, and it is necessary to borrow QPIs of other CPUs, where two A QPI of a CPU is called a hop. For example: In Figure 1, if the CPU0 in node 0 is to be connected to the CPUO in node 1, then only 4 hops of QPI are passed; however, CPU0 in node 0 To connect to CPU0 in node 2, you need to go through the three-hop QPI of 1, 2, and 3. However, the more hops, the more time it takes for the current CPU to access other CPUs, for example: The time required for the CPUO in node 0 to access the CPUO in node 2 is much larger than the CPUO access in node 0. The time required for the CPUO in node 1. From The CPU access delay is caused, so that the CPU needs to spend extra time in the operation process to wait for data transmitted by other CPUs other than the QP I, thereby reducing the operation rate of the entire partition.

In the prior art, the main cause of the above problem is that the process of dividing the node partition and interconnecting the CPU is manually performed by a technician, and in order to ensure that the partition can have a higher operation rate, the technician needs to collect the system. Monitor the data and analyze the preferred partitioning scheme based on your own experience and knowledge, which inevitably increases the workload of the staff. For some systems with unstable operation, the partition needs to be modified frequently. This requires the technician to manually monitor the operation of the system and manually modify the partition at any time. This requires a large amount of time and workload for the technician, and increases the system. Operating costs. Summary of the invention

The embodiment of the invention provides a method, a device and a server for dividing a node partition, which can automatically analyze and obtain a partitioning scheme according to the topology structure of the system, which can avoid manual division of the partition by the technician, thereby reducing the time consumption of the technician and The workload, which reduces the operating costs of the system.

In order to achieve the above object, the embodiment of the present invention adopts the following technical solutions:

In one aspect, an embodiment of the present invention provides a method for partitioning a node partition, including: acquiring a topology of a system and a number of nodes participating in the partition, the system comprising at least three nodes, each node including at least two CPUs;

Determining connection information according to the topology structure of the system, where the connection information includes: a connection relationship between each CPU and other CPUs in the system;

Determining a partitioning scheme according to the number of nodes participating in the partition and the connection information, wherein the number of nodes in the partitioning scheme is the number of nodes of the participating partitioning.

In another aspect, an embodiment of the present invention provides an apparatus for dividing a node partition, including: a topology extraction module, configured to acquire a topology structure of the system, where the system includes at least three nodes, each node includes at least two CPU ; a quantity determining module, configured to acquire the number of nodes participating in the partition;

a topology analysis module, configured to determine connection information according to a topology structure of the system, where the connection information includes: a connection relationship between each CPU in the system and another CPU; an analysis module, configured to The number of nodes participating in the partition and the connection information determine a partitioning scheme, wherein the number of nodes in the partitioning scheme is the number of nodes of the participating partitions.

In still another aspect, an embodiment of the present invention provides a server for partitioning a node partition, including: a processor, a memory, and an input device, where:

The processor acquires, from the memory and/or the input device, a topology of the system and a number of nodes participating in the partition, the system includes at least three nodes, each node includes at least two CPUs, and then Determining the connection structure of the system, the connection information includes: a connection relationship between each CPU in the system and other CPUs, and finally determining according to the number of nodes participating in the partition and the connection information a partitioning scheme, wherein the number of nodes in the partitioning scheme is the number of nodes of the participating partitioning.

The method, device and server for dividing a node partition provided by the embodiment of the invention can automatically analyze and obtain a partitioning scheme with the least number of interconnected channels according to the topology structure of the system, thereby reducing the workload of the technician and reducing the operation of the system. cost. DRAWINGS

In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the embodiments will be briefly described below. It is obvious that the drawings in the following description are only some embodiments of the present invention. Other drawings may also be obtained from those of ordinary skill in the art in light of the inventive work. 1 is a schematic structural diagram of a specific example in the prior art; FIG. 2a is a flowchart of a method for partitioning a node according to Embodiment 1 of the present invention; FIG. 2b is a schematic structural diagram of a specific example according to an embodiment of the present invention; 2a is a schematic structural diagram of another specific example according to an embodiment of the present invention; FIG. 2 is a schematic structural diagram of another embodiment of the present invention; FIG. 3a is a schematic diagram of partitioning a node partition according to Embodiment 2 of the present invention; FIG. 3b is a flowchart of another method for partitioning a node according to Embodiment 2 of the present invention; FIG. 3c is a flowchart of another method for partitioning a node according to Embodiment 2 of the present invention; FIG. 3c2 is a schematic structural diagram of another embodiment of the present invention; FIG. 3c is a schematic structural diagram of still another embodiment of the present invention; FIG. FIG. 3 is a flowchart of a method for partitioning a node partition according to Embodiment 2 of the present invention; FIG. 3e2 is a further partition node partition according to Embodiment 2 of the present invention; Method flow chart; FIG. 3f is a flow chart of another method for dividing a node partition according to Embodiment 2 of the present invention; FIG. FIG. 3g is a flowchart of another method for dividing a node partition according to Embodiment 2 of the present invention; FIG. 3g is a flowchart of a specific example provided by Embodiment 2 of the present invention; FIG. 3 is a schematic flowchart of a method for partitioning a node according to Embodiment 2 of the present invention; FIG. 3 is a schematic structural diagram of a specific example provided by Embodiment 2 of the present invention; FIG. 5 is a schematic structural diagram of an apparatus for partitioning a node partition according to Embodiment 4 of the present invention; FIG. 5b is a schematic structural diagram of another apparatus for partitioning a node partition according to Embodiment 4 of the present invention; FIG. 5 is a structural diagram of another apparatus for dividing a node partition according to Embodiment 4 of the present invention;

FIG. 5 is a schematic structural diagram of another apparatus for dividing a node partition according to Embodiment 4 of the present invention; FIG. 5 e is a schematic structural diagram of another apparatus for partitioning a node partition according to Embodiment 4 of the present invention; FIG. FIG. 5 is a schematic structural diagram of another apparatus for dividing a node partition according to Embodiment 4 of the present invention; FIG.

FIG. 6 is a schematic structural diagram of a server according to Embodiment 5 of the present invention. detailed description

The technical solutions in the embodiments of the present invention are clearly and completely described in the following with reference to the accompanying drawings in the embodiments of the present invention. It is obvious that the described embodiments are only a part of the embodiments of the present invention, but not all of the embodiments. All other embodiments obtained by a person of ordinary skill in the art based on the embodiments of the present invention without creative efforts are within the scope of the present invention.

In order to make the advantages of the technical solutions of the present invention clearer, the present invention will be described in detail below with reference to the accompanying drawings and embodiments.

Example 1

The embodiment of the present invention provides a method for dividing a node partition, as shown in FIG. 2a, including: It should be noted that the method flow of the embodiment of the present invention may be performed by a server in the system that performs system supervision, for example: In the technology, the workstation as the management server in the communication system can undertake the function of monitoring the operation and scheduling of the entire communication system, and the technician can Get the entire communication network running at any time through the workstation.

5201. The server acquires the topology structure of the system and the number of nodes participating in the partition.

The system includes at least three nodes, each of which includes at least two CPUs.

In this embodiment, the server can obtain the system topology structure by using existing technical means, and the system topology includes the number and connection relationship of each network element, for example: the communication system of the base station-mobile terminal is set up. Afterwards, devices such as routers, gateways, switches, servers, and transmitting antennas are connected by physical cables according to the topology map provided by the operator (including the number of various devices and connection relationships), and this extension is The Pak structure map is stored in the server so that the server can monitor the operation of each node in the system according to the topology structure diagram while the entire system is running.

Further, in this embodiment, the server may acquire the number of nodes participating in the partition, and the number of the nodes may be input into the server by the technical personnel through the input device, or may be automatically calculated by the server according to the specific running state.

5202. Determine connection information according to the topology of the system.

The connection information includes: a connection relationship between each CPU in the system and other CPUs.

In this embodiment, the CPUs can be connected through the QP I, and the server can directly obtain the connection relationship between the CPUs from the topology structure, for example: In practical applications, the connection relationship between the CPUs is provided by the operator. In the topology diagram, the server can be obtained from the topology map.

S203. Determine a partitioning scheme according to the number of nodes participating in the partition and the connection information. The number of nodes in the partitioning scheme is the number of nodes in the participating partition. It should be noted that, in practical applications, each device of the system is often connected through physical lines (such as data lines, coaxial cables, etc.), and the server is also connected to the port between the CPUs through corresponding physical lines, thereby being physically The CPUs are actually connected, but in the actual operation of the system, not all physical lines are enabled. For example: Figure 2b shows the three node topologies including node 0, node 1, and node 2. Figure, and in each node The CPUs are all connected by corresponding physical lines, wherein 2, 4 may be the cross-links connected between the nodes shown in Figure 2c; 1, 3 may be the internal connections of the single nodes shown in Figure 2d 2 Bridge wiring for the CPU port.

The server can enable this physical line by constructing QP I on a physical line, so that the information exchange between the two CPUs can be realized. For example, the server can construct QP I on the two physical lines 1, 4 The information interaction between CPU1 in node 0 and CPU0 in node 1 is implemented.

The method for partitioning node partitions provided by the embodiments of the present invention can automatically analyze and obtain the partitioning scheme with the least number of interconnected channels according to the topology of the system, thereby reducing the workload of the technicians and reducing the operating cost of the system.

Example 2

An embodiment of the present invention provides a method for dividing a node partition. As shown in FIG. 3a, the method includes: S301. The server acquires a topology structure of the system.

S302a, obtaining running state information of the system.

In this embodiment, the server may obtain the running state information of the system by using technical means well known to those skilled in the art, and use the running state information as a parameter to analyze and determine the number of nodes participating in the partition, for example:

The running status information may include an average load value of the CPU in the system, and the server acquires the load value of each CPU in the system, and calculates the average load value. For example: The server can get the load value of a total of 6 CPUs among the 3 nodes shown in Figure 2b and calculate the average load value.

S 303a, detecting an interval in which the average load value is located, and determining a preset value corresponding to the interval as the number of nodes of the participating partition.

In this embodiment, the server may determine the number of nodes participating in the partition by using threshold analysis according to the running state information, for example:

A series of consecutive intervals can be pre-stored in the server, and each interval corresponds to one pre- Set value, the corresponding value of the interval where the average load value is located is the number of nodes participating in the partition, and the ratio is:

Table I

Table 1 shows the correspondence between the continuous interval in the server and the preset value. If the average load value is 6, the server will use 3 as the number of nodes participating in the partition, that is, divide the total of 3 nodes into one. In the partition.

Similarly, the server can also determine the number of nodes participating in the partition according to other running status information. For example, the server can use the data traffic as the running status information, and determine the number of nodes participating in the partition by using a threshold analysis method such as described in S 303a.

In this embodiment, the server can automatically analyze the number of nodes participating in the partition according to the running state of the system by executing the processes of S 302 a to S 303 a , thereby eliminating the manual analysis and calculating the nodes participating in the partition by those skilled in the prior art. The number of processes reduces the workload of the technicians, increases the automation of the system, reduces labor costs, and ultimately reduces operating costs.

Parallelly, as shown in FIG. 3b, in the embodiment, the manner in which the server determines the number of nodes participating in the partition further includes S302b.

S302b, receiving quantity information, and acquiring the number of nodes of the participating partition according to the quantity information.

In this embodiment, the server may receive the quantity information input by the technician, and acquire the number of nodes participating in the partition according to the quantity information. That is, in this embodiment, the number of nodes participating in the partition may be determined and input by a technician responsible for system maintenance, and the server automatically generates a partitioning scheme according to the quantity information input by the technician. It should be noted that, in S 302 b, the technician only needs to input the number of nodes participating in the partition, and does not need to manually determine the same as in the prior art. Which servers participate in the partition, the process of determining the partitioning scheme is still performed automatically by the server, and does not require manual participation. E.g:

The server can display an input interface on the display. The technician can enter the value into the server through the input device. The server processes the value entered by the technician as quantity information and determines the number of nodes participating in the partition.

Further, S 302a to S 303a and S 302b may be executed in a certain order. For example, the server may first execute the processes of S 302a to S 303a, and automatically obtain a preset value by using the interval comparison method, and then The preset value display is displayed in the input interface on the display for reference by the technician and receives the quantity information input by the technician.

S304. Determine connection information according to a topology of the system.

S305. Determine a partitioning scheme according to the number of nodes participating in the partition and the connection information. The number of nodes in the partitioning scheme is the number of nodes in the participating partition. Also, each CPU in the system is connected to other CPUs via the fast interconnect channel QP I.

Further, in this embodiment, as shown in FIG. 3c, the S305 may include:

53051, determine the total number of nodes M that make up the system.

53052, obtaining a partition combination according to M and the number of nodes N participating in the partition. Wherein, N nodes are included in each component zone combination, and the N nodes are part of M nodes constituting the system, and each partition combination includes nodes different from other zone combinations. .

That is, the server can obtain the partition combination of all possible servers when the number of nodes participating in the partition is N. For example: As shown in Figure 2b, there are a total of 3 nodes in the system, and the number of nodes participating in the partition is 2, then C ₃ ² = 3 partition combinations are obtained, namely node 0 - node 1, node 0 - node 2, node 1 node 2 .

5305 3: Obtain a maximum QP I hop count in each partition combination according to the connection information. In this embodiment, the server may obtain the maximum number of QPI hops in each partition combination by means of an existing exhaustive algorithm or the like.

For example: Figure 3cl, Figure 3c2, Figure 3c3 are the three partitioning schemes in Figure 2b, where:

Figure 3cl is the partition combination of node 0-node 1, the maximum QPI hop count is 2, for example: CPU0 in node 0 is connected to CPU1 in node 1 through 1 and 7 second hop;

Figure 3c2 is the partition combination of node 0 - node 2, the maximum QPI hop count is 3, for example: CPU0 in node 0 is connected to CPU0 in node 2 through 1, 2 and 3 three hops;

Figure 3c3 shows the partition combination of node 1 and node 2. The maximum number of QPI hops is 3. For example, CPU1 in node 0 connects to CPU1 in node 2 through 6, 5, and 3 triple hops.

S3054: Determine a partition combination that minimizes a maximum QPI hop count as the partitioning scheme.

For example, the maximum QPI hop count of the partition combination shown in Figure 3cl is 2, the maximum QPI hop count shown in Figure 3c2 is 3, and the partition Q1 and the partition combination shown in Figure 3c3 have a maximum QPI hop count of 3. The partition combination shown in Fig. 3cl is determined as the partition scheme.

In this embodiment, by executing the processes of S3051 to S3054, the server can automatically analyze and obtain an optimal partitioning scheme by using a computer algorithm such as an exhaustive method, thereby avoiding the influence of subjective factors when manually analyzing the partitioning scheme, and realizing The server automatically partitions the servers in the system with the most preferred scenario partition. And because the maximum QPI hop count in the most preferred scheme partition is the smallest, the computational rate reduction caused by the excessive number of QPI hops between the CPUs is reduced, thereby reducing the performance degradation of the system.

Further, as shown in FIG. 3d, the method further includes:

S306. If there are at least two combinations of partitions with the smallest maximum number of QPI hops, it is detected whether there is identification information.

In this embodiment, there are multiple combinations of partitions with the smallest maximum number of QPI hops, for example: the maximum QPI hop count of partition combination 1 is 2, the maximum QPI hop count of partition combination 2 is 2, and the maximum partition combination 3 is The QPI hop count is 4, that is, the maximum QPI hop count of partition combination 1 and partition combination 2 is the same and both are minimum. It should be noted that the identification information in this embodiment may be pre-stored in the server and automatically obtained by the server. For example, the storage device of the server may pre-store the number of the server as the identification information, and when there is more than one maximum QP I When the partition with the smallest number of hops is combined, the server automatically extracts the number of the server, and arranges the numbers in the order of small to large, and then determines the partition combination of the highest number of the included server as the partition scheme, for example: partition combination 1 includes node 0 and node 1; partition combination 2 includes node 0 and node 2, and it is obvious that server numbers 0 and 1 in partition combination 1 are higher than server numbers 0 and 2 in partition combination 1, The server can determine partition combination 1 as a partitioning scheme.

It should also be noted that the identification information can also be input by a technician through an input device, for example:

The server can display an input interface on the display screen, and the partition interface 1 and the partition combination 2 are displayed on the input interface, and the technician can input the identification information through the input device, and the identification information corresponds to the partition combination 1 or the partition combination 2, so as to facilitate The server selects one of the partition combination 1 and the partition combination 2 as a partitioning scheme based on the identification information.

S 307a, if the identification information exists, determining the partition combination specified by the identification information as the partitioning scheme.

S 307b , if there is no identification information, selecting a partition combination from the obtained partition combination with the smallest maximum QPI hop count is determined as the partition scheme.

In this embodiment, the server can automatically select a partition combination and determine the partition scheme without the identification information. The specific implementation manner in which the server performs automatic selection may be any manner known to those skilled in the art. For example, the server may use the partition combination with the smallest maximum number of QPI hops obtained first as the partitioning scheme.

Further, after determining the partitioning scheme, as shown in FIG. 3el, the method further includes:

S 308a , detecting whether a trigger information is received.

In this embodiment, after the server automatically generates the partitioning scheme, the partitioning may be automatically performed according to the partitioning scheme, and the specific partitioning manner may be a manner well known to those skilled in the art. And the trigger information can be input by the technician through the input device to trigger the execution process of the partition. S309a. If the trigger information is received, the system is partitioned according to the determined partitioning scheme.

It should be noted that, in this embodiment, after receiving the trigger information input by the technician, the server may automatically perform partitioning according to the partitioning scheme, and the process of the entire partition does not need to be manually operated, and the technician only starts the process of triggering the partition. carried out.

If no trigger information is received, no processing is performed.

It should be noted that the server may perform S308a to S309a, or S308b after the execution of S305; or may perform S308a to S309a after executing S307a or S307b as shown in FIG. 3e2, or as shown in FIG. 3f2, after execution is completed. S307a or S307b and S308b, that is, the server may perform S308a to S309a, or S308b after determining the partitioning scheme.

Parallel, as shown in Figure 3, also includes:

S308b, partitioning the system according to the determined partitioning scheme.

That is, the technician is not required to input the trigger information, and the server automatically performs the partitioning process immediately after determining the partitioning scheme.

Further, after the system is partitioned, the method further includes:

First, as shown in Figure 3g,

S310a, obtaining a current average load value in the system.

S311a, if the current average load value is greater than the maximum value of the interval, the partition combination with the smallest maximum QPI hop count among the C^ ⁺¹ partition combinations is determined as a new partition scheme, and according to the new partition scheme Partition the system.

For example: As shown in Table 1, if the average load value of the CPU reacquired by the server is increased from the original 6 to 12, the default value is 3.

If the current average load value is not greater than the maximum value of the interval, no processing is performed. That is, the server needs to add a node to the partition. The server can re-determine the partition combination with the smallest maximum QPI hop count as the new partition scheme in the c^ ⁺¹ partition combination.

Further, the new partitioning scheme acquired by the server may include nodes in the partition currently used by the system. That is, the server can join the server in the partition where the system is already running, so that the server The new partition includes all the servers in the partition that the original system has been running and the newly joined servers. For example: As shown in Figure 3g l, there are two nodes, Node 0 and Node 1, in the partition currently used by the system. If the server needs to add another node to the partition currently used by the system, the server can pass the exhaustive method. The technical means obtains C^ ⁺¹ partition combinations, and ^{selects the} partition combination of the nodes in the partition currently used by the system (node 0 and node 1 in this example) in the C^ ⁺¹ partition combination. And selecting the combination of partitions with the smallest maximum QPI hop count among the partition combinations of the nodes in the partition currently used by the system as the new partition scheme. Where M is the total number of nodes that make up the system, and N is the number of nodes in the partition currently used by the system. The new partitioning scheme includes node 0, node 1, and node 2. Then the server adds node 2 to the partition currently used by the system, and adds node 2, node 0, and node 1 by adding 5 lines, 3, and 2 lines. Connected, the new partitioning scheme can be implemented, and in the process of implementing the new partitioning scheme, since there is no need to change the connection relationship between the node and the server in the original partition, there is no need to interrupt the partition currently used by the system. Business.

Second, as shown in Figure 3h,

S 310b , obtaining the current average load value in the system.

S 311b, if the current average load value is less than the minimum value of the interval, determining a partition combination in which the maximum QP I hop count is the smallest in the C^ ^-1 partition combination, as a new partition scheme, and according to the new The partitioning scheme partitions the system.

For example: As shown in Table 1, if the average load value of the CPU reacquired by the server is reduced from the original 12 to 6, the default value is reduced from 3 to 2.

That is, the server needs to reduce one node to the partition. The server can re-determine the partition combination with the smallest maximum QPI hop count as the new partition scheme in the c^ ^-1 partition combination.

Further, the partition currently used by the system may include nodes in the new partitioning scheme. That is, the new partitioning scheme acquired by the server may include a part of the servers in the partition currently used by the system, and the server may reduce the server in the partition in which the system has already run. For example: As shown in Figure 3hl, there are three nodes in node 0, node 1, and node 2 in the partition currently used by the system. If the server needs to reduce one node in the partition currently used by the system, the server Can be obtained through technical means such as exhaustive law

The partition combination is selected from the partition combination of the nodes in the partition currently used by the system (in this example, any two of node 0, node 1 and node 2), and has the partition currently used by the system. The partition combination in which the maximum QP I hop count is selected among the partition combinations of the nodes in the node is determined as the new partition scheme. Where M is the total number of nodes that make up the system, and N is the number of nodes in the partition currently used by the system. The new partitioning scheme includes node 0 and node 1. Then the server subtracts node 2 from the partition currently used by the system, and closes the lines connecting node 2 with node 0 and node 1 by 5, 3, and 2 lines. , the new partitioning scheme can be implemented, and in the process of implementing the new partitioning scheme, since it is not necessary to change the server other than the subtracted server in the original partition (in this example, it is the node 0, node 1) The connection relationship between them, so that there is no need to interrupt the traffic on the partition currently used by the system.

The method for partitioning node partitions provided by the embodiments of the present invention can automatically analyze and obtain the partitioning scheme with the least number of interconnected channels according to the topology of the system, thereby reducing the workload of the technicians and reducing the operating cost of the system. Moreover, the influence of subjective factors in manually analyzing the partitioning scheme is avoided, so that the server in the system is partitioned by the most preferable scheme partitioning, thereby reducing the operation rate reduction caused by improper partitioning, thereby reducing system performance degradation.

Example 3

An embodiment of the present invention provides a device for dividing a node partition, as shown in FIG. 4, including: a topology extraction module 41, configured to acquire a topology structure of the system.

The quantity determining module 42 is configured to obtain the number of nodes participating in the partition.

The topology analysis module 43 is configured to determine connection information according to the topology of the system. The connection information includes: a connection relationship between each CPU in the system and other CPUs.

The analyzing module 44 is configured to determine a partitioning scheme according to the number of nodes participating in the partition and the connection information.

Wherein, the number of nodes in the partitioning scheme is the number of nodes participating in the partition. The device for dividing a node partition provided by the embodiment of the present invention can automatically analyze and obtain a partitioning scheme with the minimum number of interconnected channels according to the topology structure of the system, thereby reducing the workload of the technician and reducing the operating cost of the system. .

Example 4

An embodiment of the present invention provides a device for dividing a node partition, as shown in FIG. 5a, including: a topology extraction module 51, configured to acquire a topology structure of the system.

The system includes at least three nodes, each of which includes at least two CPUs. The quantity determining module 52 is configured to acquire the number of nodes participating in the partition.

As shown in FIG. 5b, the quantity determining module 52 includes:

The state detecting unit 521 a is configured to acquire running state information of the system, where the running state information includes an average load value of a CPU in the system.

The quantity analysis unit 522a is configured to acquire the number of nodes of the participating partition according to the average load value.

The quantity determining unit 522a is further configured to detect that the area where the average load value is located is juxtaposed. As shown in FIG. 5c, the quantity determining module 52 may further include: a quantity information receiving unit 521b. And a number of nodes used to acquire the participating partitions according to the quantity information.

The topology analysis module 5 3 is configured to determine connection information according to the topology structure of the system. The connection information includes: The connection relationship between each CPU in the system and other CPUs. The analyzing module 54 is configured to determine a partitioning scheme according to the number of nodes participating in the partition and the connection information.

It should be noted that the number of nodes in the partitioning scheme is the number of nodes participating in the partition. The analysis module 54 can include:

A total amount determining unit 541, configured to determine the total number of nodes constituting the system^

Among them, each CPU in the system communicates with other CPUs through the fast interconnect channel QP I Connected,

a grouping unit 542, configured to acquire a seed partition combination according to M and the number N of nodes participating in the partition, where N nodes are included in each component area combination, and the N nodes are M nodes constituting the system Part of it, and each partition combination includes nodes that are different from other partition combinations.

The statistic unit 543 is configured to obtain the maximum QPI hop count in each partition combination according to the connection information.

The selecting unit 544 determines a partition combination for minimizing the maximum QPI hop count as the partition scheme.

Further, as shown in FIG. 5e, the analyzing module 54 further includes:

The identification information detecting unit 545 is configured to detect whether there is identification information if there are at least two combinations of partitions with the smallest maximum number of QPI hops.

The selecting unit 544 is further configured to determine, if the identification information exists, a partition combination specified by the identification information as the partitioning scheme.

Optionally, in this embodiment, as shown in FIG. 5f, the apparatus may further include: a partitioning module 55, configured to partition the system according to the determined partitioning scheme after determining the partitioning scheme.

Further, in this embodiment, as shown in FIG. 5g, the apparatus further includes: a receiving module 56, configured to detect whether the trigger information is received after determining the partitioning scheme. The partitioning module 55 is further configured to: if the trigger information is received, partition the system according to the determined partitioning scheme.

Further, in this embodiment, the method further includes:

The state detecting unit 521a is further configured to obtain a current average load value in the system after partitioning the system.

The selecting unit 522a is further configured to determine, as the new partitioning scheme, the partition combination in which the maximum QPI hop count is the smallest among the C^ ⁺¹ partition combinations, if the current average load value is greater than the maximum value of the interval. The partitioning module 55 is further configured to partition the system according to the new partitioning scheme. Among them, the new partitioning scheme includes the nodes in the partition currently used by the system.

Optionally, in this embodiment, the method further includes:

The state detecting unit 521 a is further configured to obtain a current average load value in the system after partitioning the system.

The selecting unit 522a is further configured to determine, as the new partitioning scheme, a partition combination in which the maximum QP I hop count is the smallest among the C^ ^-1 partition combinations, if the current average load value is less than the minimum value of the interval. .

The partitioning module 55 is further configured to partition the system according to the new partitioning scheme. The partition currently used by the system includes nodes in the new partition scheme.

The device for dividing a node partition provided by the embodiment of the present invention can automatically analyze and obtain a partitioning scheme with the minimum number of interconnected channels according to the topology structure of the system, thereby reducing the workload of the technician and reducing the operating cost of the system. . Moreover, the influence of subjective factors in the manual analysis of the partitioning scheme is avoided, so that the servers in the system are partitioned by the most preferable scheme partitioning, thereby reducing the operation rate reduction caused by improper partitioning, thereby reducing the system performance degradation.

Example 5

The embodiment of the present invention provides a server for partitioning a node partition. As shown in FIG. 6, the server includes: a processor 61, a memory 62, and an input device 63, where:

The processor 61 acquires the topology structure of the system and the number of nodes participating in the partition from the memory 62 or the input device 63; and further determines connection information according to the topology structure of the system; and finally according to the participating partition The number of nodes and the connection information determine the partitioning scheme.

It should be noted that the system includes at least three nodes, and each node includes at least two CPUs. The connection information includes: The connection relationship between each CPU in the system and other CPUs. The number of nodes in the partitioning scheme is the number of nodes participating in the partition.

The number of nodes that the processor 61 acquires the participating partitions includes: the processor 61 Obtaining running state information of the system, where the running state information includes an average load value of the CPU in the system, and acquiring the number of nodes participating in the partition according to the average load value.

Further, the obtaining, by the processor 61, the number of nodes participating in the partition according to the average load value includes: the processor 61 detecting an interval in which the average load value is located, and acquiring, from the memory 62, the The preset corresponding to the interval, and determining the preset value as the number of nodes participating in the partition.

Optionally, the processor 61 may further receive the quantity information by using the input device 63, and obtain the number of nodes participating in the partition according to the quantity information.

Specifically, in this embodiment, the determining, by the processor 61, the partitioning scheme according to the number of nodes participating in the partition and the connection information may include:

The processor 61 determines the total number M of nodes constituting the system, and obtains a seed partition combination according to M and the number of nodes N participating in the partition, wherein N nodes are included in each component area combination, and this N The nodes are part of the M nodes constituting the system, and each of the partition combinations includes nodes that are different from other partition combinations, and then obtain the largest of each partition combination according to the connection information. The QPI hop count, and finally the partition combination that minimizes the maximum QP I hop count is determined as the partition scheme.

It should be noted that each CPU in the system is connected to other CPUs through the quick interconnect channel QPI.

The processor 61 detects whether there is identification information, and the identification information is acquired by the processor 61 through the input device 63, or by the The processor 61 is obtained from the memory 62. If the identification information exists, the processor 61 determines the partition combination specified by the identification information as the partition scheme.

On the other hand, after determining the partitioning scheme, the processor 61 may partition the system according to the determined partitioning scheme.

In still another aspect, after determining the partitioning scheme, the processor 61 detects whether trigger information is received through the input device 63. If the input device 63 receives the trigger information, then The processor 61 partitions the system according to the determined partitioning scheme.

The server for dividing the node partition described in this embodiment may further include:

The processor 61 obtains the current average load value in the system after partitioning the system. If the current average load value is greater than the maximum value of the interval, the processor 61 determines, as a new partitioning scheme, a partition combination in which the maximum QP I hop count is the smallest among the C^ ⁺¹ partition combinations, and according to the The new partitioning scheme partitions the system. Among them, the new partitioning scheme includes the nodes in the partition currently used by the system.

If the current average load value is less than the minimum value of the interval, the processor 61 determines, as a new partitioning scheme, a partition combination in which the maximum QP I hop count is the smallest among the C^ ^-1 partition combinations, and according to the The new partitioning scheme partitions the system. The partition currently used by the system includes nodes in the new partition scheme.

The server dividing the node partition provided by the embodiment of the present invention can automatically analyze and obtain the partitioning scheme with the least number of interconnected channels according to the topology structure of the system, thereby reducing the workload of the technician and reducing the operating cost of the system. Moreover, the influence of subjective factors in manually analyzing the partitioning scheme is avoided, so that the server in the system is partitioned by the most preferable scheme partitioning, thereby reducing the operation rate reduction caused by improper partitioning, thereby reducing the system performance degradation. A person skilled in the art can understand that all or part of the process of implementing the above embodiment method can be completed by a computer program to instruct related hardware, and the program can be stored in a computer readable storage medium. In execution, the flow of an embodiment of the methods as described above may be included. The storage medium may be a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

The above is only the specific embodiment of the present invention, but the scope of the present invention is not limited thereto, and any change or replacement that can be easily conceived by those skilled in the art within the technical scope of the present invention is All should be covered by the scope of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

claims

1. A method for dividing node partitions, which is characterized by including:

The server obtains the topology structure of the system and the number of nodes participating in the partition. The system includes at least three nodes, and each node includes at least two CPUs;

According to the topological structure of the system, connection information is determined, and the connection information includes: the connection relationship between each CPU and other CPUs in the system;

A partition plan is determined according to the number of nodes participating in the partition and the connection information, where the number of nodes in the partition plan is the number of nodes participating in the partition.

2. The method of dividing node partitions according to claim 1, wherein the server obtains the number of nodes participating in the partitions by:

The server obtains the running status information of the system, and the running status information includes the average load value of the CPU in the system;

Obtain the number of nodes participating in the partition according to the average load value.

3. The method of dividing node partitions according to claim 1, characterized in that: obtaining the number of nodes participating in the partition according to the average load value includes:

The server detects the interval in which the average load value is located, and determines the preset value corresponding to the interval as the number of nodes participating in the partition.

4. The method of dividing node partitions according to claim 1, wherein the obtaining the number of nodes participating in the partitions includes:

The server receives quantity information and obtains the number of nodes participating in the partition based on the quantity information.

5. The method of dividing node partitions according to claim 1, characterized in that each CPU in the system is connected to other CPUs through a fast interconnection channel QPI, and the number of nodes participating in the partition is determined by the number of nodes participating in the partition. and the connection information determines the partitioning scheme including:

The server determines the total number M of nodes that make up the system;

According to M and the number of nodes participating in the partition N, obtain a partition combination, where Each partition combination includes N nodes, which are part of the M nodes that make up the system, and each partition combination includes nodes that are different from other partition combinations;

According to the connection information, the maximum number of QPI hops in each partition combination is obtained; and the partition combination with the smallest maximum number of QPI hops is determined as the partition scheme.

6. The method of dividing node partitions according to claim 5, further comprising:

If there are at least two partition combinations with the smallest maximum QPI hop count, then the server detects whether there is identification information;

If identification information exists, the partition combination specified by the identification information is determined as the partition scheme.

7. The method of dividing node partitions according to claim 1, characterized in that, after determining the partition scheme, the method further includes:

The server partitions the system according to the determined partitioning scheme.

8. The method of dividing node partitions according to claim 1, characterized in that, after determining the partition scheme, the method further includes:

The server detects whether trigger information is received;

If the trigger information is received, the system is partitioned according to the determined partition plan.

9. The method of dividing node partitions according to claim 5 or 7, characterized in that, after partitioning the system, it further includes:

The server obtains the current average load value in the system;

If the current average load value is greater than the maximum value of the interval, then the partition combination with the smallest maximum QPI hop number among c^ ⁺¹ partition combinations is determined as a new partition scheme, and all partition combinations are configured according to the new partition scheme. partition the system.

10. The method of dividing node partitions according to claim 9, wherein the new partition scheme includes nodes in the partition currently used by the system.

11. The method of dividing node partitions according to claim 5 or 7, characterized in that, after partitioning the system, it further includes:

The server obtains the current average load value in the system;

If the current average load value is less than the minimum value of the interval, then the partition combination with the smallest maximum QPI hop number among C^ ^-1 partition combinations is determined as a new partition scheme, and the partition combination is determined according to the new partition scheme. The system is partitioned.

12. The method of dividing node partitions according to claim 11, characterized in that the partition currently used by the system includes nodes in the new partition scheme.

1 3. A device for dividing node partitions, characterized by including:

The topology extraction module is used to obtain the topology structure of the system. The system includes at least three nodes, and each node includes at least two CPUs;

Quantity determination module, used to obtain the number of nodes participating in the partition;

The topology analysis module is used to determine the connection information according to the topology structure of the system. The connection information includes: the connection relationship between each CPU and other CPUs in the system; the analysis module is used to determine the connection information according to the topology structure of the system. The number of nodes participating in the partition and the connection information determine a partition plan, where the number of nodes in the partition plan is the number of nodes participating in the partition.

14. The device for dividing node partitions according to claim 13, characterized in that the quantity determination module includes:

A status detection unit, used to obtain the operating status information of the system, where the operating status information includes the average load value of the CPU in the system;

A quantity analysis unit, configured to obtain the number of nodes participating in the partition according to the average load value.

15. The device for dividing node partitions according to claim 14, characterized in that the quantity analysis unit is also used to detect the interval in which the average load value is located, and set the preset value corresponding to the interval Determine the number of nodes participating in the partition.

16. The device for dividing node partitions according to claim 13, characterized in that: The quantity determination module includes:

A quantity information receiving unit, configured to obtain the number of nodes participating in the partition according to the quantity information.

17. The device for dividing node partitions according to claim 13, characterized in that each CPU in the system is connected to other CPUs through a fast interconnection channel QPI, and the analysis module includes:

The total amount determination unit is used to determine the total number M of nodes that make up the system;

A grouping unit, used to obtain a partition combination based on M and the number N of nodes participating in the partition, where each group of partition combinations includes N nodes, and these N nodes are among the M nodes that make up the system. part of , and each partition combination includes nodes that are different from other partition combinations;

A statistical unit, used to obtain the maximum number of QPI hops in each partition combination based on the connection information;

A selection unit is used to determine the partition combination with the smallest maximum QPI hop count as the partition scheme.

18. The device for dividing node partitions according to claim 17, characterized in that the analysis module further includes:

The identification information detection unit is used to detect whether there is identification information if there are at least two partition combinations with the smallest maximum QPI hop count;

The selection unit is also configured to, if identification information exists, determine the partition combination specified by the identification information as the partition scheme.

19. The device for dividing node partitions according to claim 13, characterized in that the device further includes:

The partitioning module is used to partition the system according to the determined partitioning scheme after determining the partitioning scheme.

20. The device for dividing node partitions according to claim 13, characterized in that the device further includes:

The receiving module is used to detect whether trigger information is received after determining the partition plan; The partition module is also configured to partition the system according to the determined partition scheme if the trigger information is received.

21. The device for dividing node partitions according to claim 17 or 19, further comprising:

The status detection unit is also used to obtain the current average load value in the system after partitioning the system;

The selection unit is also configured to determine the partition combination with the smallest maximum QPI hop count among C^ ⁺¹ partition combinations as a new partition scheme if the current average load value is greater than the maximum value of the interval;

The partition module is also used to partition the system according to the new partition scheme.

22. The method of dividing node partitions according to claim 21, characterized in that the new partition scheme includes nodes in the partition currently used by the system.

23. The device for dividing node partitions according to claim 17 or 19, characterized in that, after partitioning the system, it further includes:

The selection unit is also configured to determine the partition combination with the smallest maximum QPI hop count among C^ ^-1 partition combinations as a new partition scheme if the current average load value is less than the minimum value of the interval;

24. The method of dividing node partitions according to claim 23, wherein the partition currently used by the system includes nodes in the new partition scheme.

25. A device for dividing a server into partitioned servers, characterized in that it includes: a processor, a memory, and an input device, wherein:

The processor obtains the topology structure of the system and the number of nodes participating in the partition from the memory or the input device. The system includes at least three nodes, each node includes at least two CPUs, and then according to the system The topology structure of , determines the connection information, and the connection information includes: The connection relationship between each CPU and other CPUs in the system, and finally determine the partition plan according to the number of nodes participating in the partition and the connection information, where the number of nodes in the partition plan is Describe the number of nodes participating in the partition.

26. The server for dividing node partitions according to claim 25, characterized in that, the processor obtaining the number of nodes participating in the partition includes:

The processor obtains the running status information of the system, the running status information includes the average load value of the CPU in the system, and then obtains the number of nodes participating in the partition based on the average load value.

27. The server for dividing node partitions according to claim 26, characterized in that, the processor obtaining the number of nodes participating in the partition according to the average load value includes: the processor detecting where the average load value is located interval, obtain a preset value corresponding to the interval from the memory, and determine the preset value as the number of nodes participating in the partition.

28. The server for dividing node partitions according to claim 25, characterized in that, the processor obtaining the number of nodes participating in the partition includes:

The processor receives quantity information through the input device, and obtains the number of nodes participating in the partition according to the quantity information.

29. The server for dividing node partitions according to claim 25, characterized in that each CPU in the system is connected to other CPUs through a fast interconnection channel QPI, and the processor participates in the partition according to the The number of nodes and the connection information determine the partitioning scheme including:

The processor determines the total number M of nodes that make up the system, and then obtains a partition combination based on M and the number N of nodes participating in the partition, where each group of partition combinations includes N nodes, and these N The node is part of the M nodes that make up the system, and each partition combination includes nodes that are different from other partition combinations. Then, based on the connection information, the maximum QP in each partition combination is obtained. I hop count, and finally the partition combination with the smallest maximum QPI I hop count is determined as the partition scheme.

30. The server for dividing node partitions according to claim 29, further comprising:

If there are at least two partition combinations with the smallest maximum QPI hop count, the processor detects whether there is identification information. The identification information is obtained by the processor through the input device, or is obtained by the processor from all obtained from the above memory;

If identification information exists, the processor determines the partition combination specified by the identification information as the partition scheme.

31. The server for dividing node partitions according to claim 25, further comprising:

After determining the partition plan, the processor partitions the system according to the determined partition plan.

32. The server for dividing node partitions according to claim 25, further comprising:

After determining the partition scheme, the processor detects whether trigger information is received through the input device;

If the input device receives the trigger information, the processor partitions the system according to the determined partition scheme.

33. The server for dividing node partitions according to claim 29 or 31, further comprising:

After partitioning the system, the processor obtains the current average load value in the system;

If the current average load value is greater than the maximum value of the interval, the processor determines the partition combination with the smallest maximum QPI hop count among ^C U ^{+ 1} partition combinations as the new partition scheme, and based on the The new partitioning scheme partitions the system.

34. The method of dividing node partitions according to claim 33, characterized in that the new partition scheme includes nodes in the partition currently used by the system.

35. The server for dividing node partitions according to claim 29 or 31, characterized in that For, also includes:

If the current average load value is less than the minimum value of the interval, the processor determines the partition combination with the smallest maximum QPI hop count among C^ ^-1 partition combinations as a new partition scheme, and determines the partition combination according to the new partition combination. A partitioning scheme partitions the system.

36. The method of dividing node partitions according to claim 35, characterized in that the partition currently used by the system includes nodes in the new partition scheme.