Load control method and device of subnet manager, medium and electronic equipment
By monitoring the CPU load of the subnet manager and sending detection management information, the detection operation of the computing node is restricted, and the problem of excessive CPU load of the subnet manager in the event of network failure is solved, effectively adjusting the load of the subnet manager and improving the operating efficiency of the network cluster are achieved.
Patent Information
- Application Number
- CN202411975269.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-30
- Publication Date
- 2025-04-18
AI Technical Summary
In a network cluster with IB subnet technology, the subnet manager may overload or get stuck in the process of a subnet manager during a network failure due to high-frequency ICMP detection, which may cause the subnet manager to overload or get stuck, resulting in the entire network system being paralyzed.
By monitoring the CPU load information of the subnet manager, load record information is generated, and the target operation mode is determined based on the preset load threshold. Then, the detection management information is sent to the computing node to limit its detection operations on the subnet manager, thereby regulating the CPU load.
This method can efficiently adjust the CPU load of the subnet manager, reduce high load conditions, improve the operating efficiency of the subnet manager of the network cluster, and avoid the network system being paralyzed.
Smart Images

Figure CN119847759A_ABST
Abstract
Description
Technical Field
[0001] The embodiments of the present application relate to the field of computer technology, and in particular to a load control method, device, medium and electronic device for a subnet manager. Background Art
[0002] IB subnet technology has been increasingly widely used in the field of supercomputer clusters. The performance of network clusters using IB subnet technology has received increasing attention. A network cluster includes a subnet manager, a switch, a storage cluster node, and a computing node. The subnet manager controls the communication between the computing node and the storage cluster node via the switch. The subnet manager includes a master service node and a slave service node in the master-slave mode. In such a large-scale cluster based on an IB network, if one or more switches fail and the master-slave switch of the subnet manager occurs at the same time, the network connection between the computing node and the storage cluster node is disconnected. In order to provide services as soon as possible after the network is restored, each node starts to perform detection operations, such as ICMP detection. Due to the master-slave switch of the subnet manager, the path information cache is invalid. The IPoIB (Internet Protocol over IB, running the IP protocol on the IB network) driver of each storage cluster node in the network cluster will send a PathRecord request to the subnet manager to query the path information.
[0003] Taking two IB subnets as an example, during ICMP detection, if 2000+ computing nodes send ICMP detection packets to 200 storage nodes, each link (each computing node corresponds to 2 IP addresses stored, a total of 2 links) once per minute, the detection frequency is about 2000*200*2, about 13,000 times per second. For each ICMP detection, in the event of a network failure, the path query will be retried 10 times, then the path query is about 130K / s. If it is a larger-scale storage cluster, more nodes will be affected, and the instantaneous request volume will be even greater. This high-frequency ICMP message detection increases the CPU load of the subnet manager, which may cause the subnet manager process to overload or freeze, thereby paralyzing the entire network system.
[0004] The existing IB subnet management technology cannot automatically control the load of the subnet manager, and the subnet manager itself does not have a mechanism to limit this high load. When the above scenario occurs, the entire network system is paralyzed, and the storage system's external services are interrupted, which will seriously affect the stability of the storage system. The current recovery method is to manually reduce the instantaneous load of the subnet manager by restarting the computing nodes in batches, so that the subnet manager can provide services normally and ensure that the cluster network returns to normal. The operating efficiency of the subnet manager of the network cluster is low. Therefore, how to improve the operating efficiency of the subnet manager of the network cluster is of great practical significance. Summary of the invention
[0005] The embodiments of the present application provide a load control method, device, medium and electronic device for a subnet manager, which can improve the operating efficiency of the subnet manager of a network cluster.
[0006] To achieve the above purpose, the technical solution of the embodiment of the present application is implemented as follows:
[0007] In a first aspect, an embodiment of the present application provides a load control method for a subnet manager, which is applied to a storage cluster node of a network cluster based on an infinite bandwidth network; the network cluster includes a subnet manager, the storage cluster node and a computing node; the method includes:
[0008] Generate load record information of the subnet manager according to the CPU load information; the load record information includes the received CPU load information of the subnet manager; the CPU load information is sent by the subnet manager to the storage cluster node;
[0009] Determining a target operation mode of the subnet manager according to the load record information and a preset load threshold;
[0010] According to the target operation mode, detection management information is sent to the computing node, so that the computing node limits the detection operation for the subnet manager according to the detection management information.
[0011] A load control method for a subnet manager provided in an embodiment of the present application is applied to a storage cluster node of a network cluster based on an infinite bandwidth network; the network cluster includes a subnet manager, the storage cluster node and a computing node; the method includes: generating load record information of the subnet manager according to CPU load information; the load record information includes received CPU load information of the subnet manager; the CPU load information is sent by the subnet manager to the storage cluster node; determining a target operation mode of the subnet manager according to the load record information and a preset load threshold; and sending detection management information to the computing node according to the target operation mode, so that the computing node limits the detection operation for the subnet manager according to the detection management information. The method generates load record information by storing the CPU load information of the subnet manager received by the cluster node; determines the target operation mode of the subnet manager according to the load record information and a preset load threshold; sends detection management information to the computing node according to the target operation mode, so that the computing node limits the detection operation for the subnet manager according to the detection management information, and provides a load control mechanism for the subnet manager of a network cluster, which can efficiently adjust the CPU load of the subnet manager, reduce the high load of the CPU of the subnet manager, and improve the operation efficiency of the subnet manager of the network cluster.
[0012] In an optional embodiment, the target operation mode includes a normal mode and a silent mode; and determining the target operation mode of the subnet manager according to the load record information and a preset load threshold includes:
[0013] According to the load record information and the preset load threshold, determining whether the control cycle threshold CPU load information up to the current moment is all less than the load threshold;
[0014] If it is monitored that the CPU load information of the control cycle threshold as of the current moment is all less than the load threshold, the target operation mode of the subnet manager is the normal mode;
[0015] If it is monitored that the control cycle threshold CPU load information up to the current moment is greater than the load threshold, the target operation mode of the subnet manager is the silent mode.
[0016] The method of this embodiment, the target operation mode includes a normal mode and a silent mode; the target operation mode of the subnet manager is determined according to the load record information and the preset load threshold, including: judging whether the control cycle threshold CPU load information up to the current moment is less than the load threshold according to the load record information and the preset load threshold; if it is monitored that the control cycle threshold CPU load information up to the current moment is less than the load threshold, the target operation mode of the subnet manager is the normal mode; if it is monitored that the control cycle threshold CPU load information up to the current moment is greater than the load threshold, the target operation mode of the subnet manager is the silent mode, and the target operation mode can be accurately set to the normal mode and the silent mode, the normal mode characterization allows detection operations on the subnet manager, and the silent mode characterization does not allow detection operations on the subnet manager, so as to further achieve more accurate adjustment of the CPU load of the subnet manager based on the normal mode and the silent mode, reduce the high load of the CPU of the subnet manager, and improve the operation efficiency of the subnet manager of the network cluster.
[0017] In an optional embodiment, the method further includes:
[0018] If the target operation mode is the silent mode and a detection request sent by the first computing node is received, determining whether the authorization queue is full;
[0019] If the result of the judgment is that the authorization queue is not full, the first computing node is added to the authorization queue, and an authorization control time of the first computing node is set; the authorization control time is used to remove the first computing node from the authorization queue after the authorization control time is reached, so as to terminate the authorization of the first computing node;
[0020] If the result of the judgment is that the authorization queue is full, the first computing node is refused to be added to the authorization queue, and the authorization blocking time of the first computing node is set; the authorization blocking time is used to prohibit the first computing node from sending a detection request again before the authorization organization time is reached.
[0021] The method of this embodiment also includes: if the target operation mode is silent mode, and a detection request sent by the first computing node is received, then judging whether the authorization queue is full; if the result of the judgment is that the authorization queue is not full, then adding the first computing node to the authorization queue, and setting the authorization control time of the first computing node; the authorization control time is used to remove the first computing node from the authorization queue after reaching the authorization control time to terminate the authorization of the first computing node; if the result of the judgment is that the authorization queue is full, then refusing to add the first computing node to the authorization queue, and setting the authorization blocking time of the first computing node; the authorization blocking time is used to prohibit the first computing node from sending the detection request again before the authorization organization time is reached. This method provides a control mechanism for computing nodes to initiate detection requests in silent mode, which can further realize more fine-tuning the CPU load of the subnet manager based on normal mode and silent mode, reduce the high load of the CPU of the subnet manager, and take into account the performance stability of the network cluster at the same time, and improve the operation efficiency of the subnet manager of the network cluster.
[0022] In a second aspect, an embodiment of the present application further provides a load control method for a subnet manager, which is applied to a subnet manager of a network cluster based on an infinite bandwidth network; the network cluster includes the subnet manager, a storage cluster node, and a computing node; the method includes:
[0023] Monitoring the CPU load information of the subnet manager;
[0024] Sending the CPU load information of the subnet manager to the storage cluster node, so that the storage cluster node generates the load record information of the subnet manager according to the CPU load information, and determines the target operation mode of the subnet manager according to the load record information of the subnet manager and a preset load threshold, and sending detection management information to the computing node according to the target operation mode; the detection management information is used by the computing node to limit the detection operation on the subnet manager; the load record information includes the received CPU load information of the subnet manager.
[0025] A load control method for a subnet manager provided in an embodiment of the present application is applied to a subnet manager of a network cluster based on an infinite bandwidth network; the network cluster includes the subnet manager, a storage cluster node and a computing node; the method includes: monitoring the CPU load information of the subnet manager; sending the CPU load information of the subnet manager to the storage cluster node, so that the storage cluster node generates load record information of the subnet manager according to the CPU load information, and enables the storage cluster node to determine the target operation mode of the subnet manager according to the load record information of the subnet manager and a preset load threshold, and sends detection management information to the computing node according to the target operation mode; the detection management information is used by the computing node to limit the detection operation on the subnet manager; the load record information includes the received CPU load information of the subnet manager. The method monitors the CPU load information of the subnet manager through the subnet manager; sends the CPU load information of the subnet manager to the storage cluster node, so that the storage cluster node generates the load record information of the subnet manager according to the CPU load information, and determines the target operation mode of the subnet manager according to the load record information of the subnet manager and a preset load threshold, and sends detection management information to the computing node according to the target operation mode; the detection management information is used by the computing node to limit the detection operation on the subnet manager, and provides a load control mechanism for the subnet manager of the network cluster, which can efficiently adjust the CPU load of the subnet manager, reduce the high load of the CPU of the subnet manager, and improve the operation efficiency of the subnet manager of the network cluster.
[0026] In a third aspect, an embodiment of the present application further provides a load control method for a subnet manager, which is applied to a computing node of a network cluster based on an infinite bandwidth network; the network cluster includes a subnet manager, a storage cluster node and the computing node; the method includes:
[0027] Monitor the detection management information sent by the storage cluster node; the detection management information is generated by the storage cluster node according to the target operation mode and sent to the computing node; the target operation mode is determined according to the load record information and a preset load threshold; the load record information is generated according to the CPU load information; the load record information includes the CPU load information of the subnet manager received by the storage cluster node; the CPU load information is sent by the subnet manager to the storage cluster node;
[0028] If the detection management information is received, the detection operation on the subnet manager is restricted according to the received detection management information.
[0029] A load control method for a subnet manager provided in an embodiment of the present application is applied to a computing node of a network cluster based on an infinite bandwidth network; the network cluster includes a subnet manager, a storage cluster node and the computing node; the method includes: monitoring detection management information sent by the storage cluster node; the detection management information is generated by the storage cluster node according to the target operation mode and sent to the computing node; the target operation mode is determined according to load record information and a preset load threshold; the load record information is generated according to CPU load information; the load record information includes the CPU load information of the subnet manager received by the storage cluster node; the CPU load information is sent by the subnet manager to the storage cluster node; if the detection management information is received, then according to the received detection management information, the detection operation for the subnet manager is restricted. The method monitors the detection management information sent by the storage cluster node through a computing node; if the detection management information is received, the detection operation on the subnet manager is limited according to the received detection management information; wherein the detection management information is generated by the storage cluster node according to the target operation mode and sent to the computing node; the target operation mode is determined according to load record information and a preset load threshold; the load record information is generated according to CPU load information; the load record information includes the CPU load information of the subnet manager received by the storage cluster node; the CPU load information is sent by the subnet manager to the storage cluster node, and a load control mechanism of the subnet manager of a network cluster is provided, which can efficiently adjust the CPU load of the subnet manager, reduce the high load of the CPU of the subnet manager, and improve the operation efficiency of the subnet manager of the network cluster.
[0030] In an optional embodiment, the target operation mode includes a normal mode and a silent mode; the detection management information includes first detection management information and second detection management information; and limiting the detection operation for the subnet manager according to the received detection management information includes:
[0031] If the detection management information is the first detection management information, the detection operation for the subnet manager is allowed to be performed; the first detection management information is sent by the storage cluster node to the computing node when the target operation mode is the normal mode;
[0032] If the detection management information is the second detection management information, the detection operation on the subnet manager is prohibited; the second detection management information is sent by the storage cluster node to the computing node when the target operation mode is the silent mode.
[0033] In the method of this embodiment, the target operation mode includes a normal mode and a silent mode; the detection management information includes a first detection management information and a second detection management information; the detection operation for the subnet manager is limited according to the received detection management information, including: if the detection management information is the first detection management information, the detection operation for the subnet manager is allowed; the first detection management information is sent by the storage cluster node to the computing node when the target operation mode is the normal mode; if the detection management information is the second detection management information, the detection operation for the subnet manager is prohibited; the second detection management information is sent by the storage cluster node to the computing node when the target operation mode is the silent mode. In this method, the target operation mode is set to a normal mode and a silent mode, the normal mode indicates that the detection operation for the subnet manager is allowed, and the silent mode indicates that the detection operation for the subnet manager is not allowed, so as to further achieve more accurate adjustment of the CPU load of the subnet manager based on the normal mode and the silent mode, reduce the high load of the CPU of the subnet manager, and improve the operation efficiency of the subnet manager of the network cluster.
[0034] In a fourth aspect, an embodiment of the present application further provides a load control device for a subnet manager, which is applied to a storage cluster node of a network cluster based on an infinite bandwidth network; the network cluster includes a subnet manager, the storage cluster node and a computing node; the device includes:
[0035] An information management unit, configured to generate load record information of the subnet manager according to the CPU load information; the load record information includes the received CPU load information of the subnet manager; the CPU load information is sent by the subnet manager to the storage cluster node;
[0036] A mode determination unit, configured to determine a target operation mode of the subnet manager according to the load record information and a preset load threshold;
[0037] A load control unit is used to send detection management information to the computing node according to the target operation mode, so that the computing node limits the detection operation for the subnet manager according to the detection management information.
[0038] In an optional embodiment, the target operation mode includes a normal mode and a silent mode; and the mode determination unit is specifically configured to:
[0039] According to the load record information and the preset load threshold, determining whether the control cycle threshold CPU load information up to the current moment is all less than the load threshold;
[0040] If it is monitored that the CPU load information of the control cycle threshold as of the current moment is all less than the load threshold, the target operation mode of the subnet manager is the normal mode;
[0041] If it is monitored that the control cycle threshold CPU load information up to the current moment is greater than the load threshold, the target operation mode of the subnet manager is the silent mode.
[0042] In an optional embodiment, the device further includes an authorization control unit; the authorization control unit is used to:
[0043] If the target operation mode is the silent mode and a detection request sent by the first computing node is received, determining whether the authorization queue is full;
[0044] If the result of the judgment is that the authorization queue is not full, the first computing node is added to the authorization queue, and an authorization control time of the first computing node is set; the authorization control time is used to remove the first computing node from the authorization queue after the authorization control time is reached, so as to terminate the authorization of the first computing node;
[0045] If the result of the judgment is that the authorization queue is full, the first computing node is refused to be added to the authorization queue, and the authorization blocking time of the first computing node is set; the authorization blocking time is used to prohibit the first computing node from sending a detection request again before the authorization organization time is reached.
[0046] In a fifth aspect, an embodiment of the present application further provides a load control device for a subnet manager, which is applied to a subnet manager of a network cluster based on an infinite bandwidth network; the network cluster includes the subnet manager, a storage cluster node and a computing node; the device includes:
[0047] An information monitoring unit, used to monitor the CPU load information of the subnet manager;
[0048] An information reporting unit is used to send the CPU load information of the subnet manager to the storage cluster node, so that the storage cluster node generates the load record information of the subnet manager according to the CPU load information, and the storage cluster node determines the target operation mode of the subnet manager according to the load record information of the subnet manager and a preset load threshold, and sends detection management information to the computing node according to the target operation mode; the detection management information is used by the computing node to limit the detection operation on the subnet manager; the load record information includes the received CPU load information of the subnet manager.
[0049] In a sixth aspect, an embodiment of the present application further provides a load control device for a subnet manager, which is applied to a computing node of a network cluster based on an infinite bandwidth network; the network cluster includes a subnet manager, a storage cluster node and the computing node; the device includes:
[0050] A management information monitoring unit, used for monitoring the detection management information sent by the storage cluster node; the detection management information is generated by the storage cluster node according to the target operation mode and sent to the computing node; the target operation mode is determined according to the load record information and a preset load threshold; the load record information is generated according to the CPU load information; the load record information includes the CPU load information of the subnet manager received by the storage cluster node; the CPU load information is sent by the subnet manager to the storage cluster node;
[0051] The receiving information response unit is used to limit the detection operation for the subnet manager according to the received detection management information if the detection management information is received.
[0052] In the seventh aspect, an embodiment of the present application further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the load control method of the subnet manager of any one of the first to third aspects is implemented.
[0053] In the eighth aspect, an embodiment of the present application further provides an electronic device, comprising a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and when the computer program is executed by the processor, the processor implements any one of the load control methods of the subnet manager in the first to third aspects.
[0054] The technical effects brought about by any one of the implementation methods in the fourth to eighth aspects can refer to the technical effects brought about by the corresponding implementation methods in the first to third aspects, and will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS
[0055] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.
[0056] Figure 1 A schematic diagram of a flow chart of a load control method of a subnet manager provided in an embodiment of the present application;
[0057] Figure 2 A schematic diagram of a flow chart of determining a target operating mode of a load control method of a subnet manager provided in an embodiment of the present application;
[0058] Figure 3 A schematic diagram of a flow chart of authorization management in silent mode of a load control method of a subnet manager provided in an embodiment of the present application;
[0059] Figure 4 A flow chart of another load control method of a subnet manager provided in an embodiment of the present application;
[0060] Figure 5 A schematic flow chart of a load control method for a third subnet manager provided in an embodiment of the present application;
[0061] Figure 6 A flow chart of a load control method for a subnet manager provided in an embodiment of the present application, which limits the detection operation of the subnet manager;
[0062] Figure 7 One of the structural schematic diagrams of a load control device of a subnet manager provided in an embodiment of the present application;
[0063] Figure 8 A second structural diagram of a load control device of a subnet manager provided in an embodiment of the present application;
[0064] Fig. 9 A structural schematic diagram of another load control device of a subnet manager provided in an embodiment of the present application;
[0065] Fig.10 A schematic diagram of the structure of a load control device of a third subnet manager provided in an embodiment of the present application;
[0066] Fig.11 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION
[0067] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present application.
[0068] It should be noted that the terms "including" and "having" and their variations involved in the documents of this application are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.
[0069] The following are explanations of some of the terms that appear in the text:
[0070] (1) Infiniband (IB): Infiniband is an open standard network interconnection technology with high bandwidth, low latency and high reliability. This technology is defined by IBTA (InfiniBand Trade Association) and has been widely used in the field of supercomputer clusters.
[0071] (2) Open Subnet Manager (opensm): The subnet manager is used for configuration and management such as IB network discovery and maintenance.
[0072] (3) Stor Master: Stor Master refers to the main service node of the storage cluster; it can also be regarded as the main service for load monitoring and control, running on the storage cluster node. Stor Master is responsible for collecting, recording and analyzing the load information reported by Stor SM Agent, and making detection actions based on the load information.
[0073] (4) Stor Master backup: Stor Master backup refers to the storage cluster's main service backup node; it can also be seen as a load monitoring and control main backup service, running on the storage cluster node. When the main service node Stor Master fails, it automatically switches to the main service backup node Stor Master backup to provide services. In some embodiments, when the Stor Master service fails, the remaining cluster nodes Stor Master backup will elect a node's StorMaster backup to run as the Stor Master.
[0074] (5) Stor Client Agent: Stor Client Agent is a storage computing node agent service. It is an agent service running on the storage computing node (client node) and is used to execute instructions issued by the Stor Master, report client information and requests, etc.
[0075] (6) Stor SM Agent: Stor SM Agent refers to the agent service running on the subnet manager (including OpenSM Master and OpenSM Slave). Stor SM Agent can be regarded as the IB subnet management node agent service, which is responsible for monitoring and reporting the OpenSM load to the Stor Master. The OpenSM Master and the corresponding OpenSM Slave constitute a service group; a service group is only responsible for monitoring and reporting the load of an IB subnet.
[0076] (7) ICMP (Internet Control Message Protocol): ICMP is a sub-protocol of the TCP / IP protocol suite, used to transmit control messages between IP hosts and routers.
[0077] IB subnet technology has been increasingly widely used in the field of supercomputer clusters. The performance of network clusters using IB subnet technology has received increasing attention. A network cluster includes a subnet manager, a switch, a storage cluster node, and a computing node. The subnet manager controls the communication between the computing node and the storage cluster node via the switch. The subnet manager includes a master service node and a slave service node in the master-slave mode. In such a large-scale cluster based on an IB network, if one or more switches fail and the master-slave switch of the subnet manager occurs at the same time, the network connection between the computing node and the storage cluster node is disconnected. In order to provide services as soon as possible after the network is restored, each node starts to perform detection operations, such as ICMP detection. Due to the master-slave switch of the subnet manager, the path information cache is invalid. The IPoIB (Internet Protocol over IB, running the IP protocol on the IB network) driver of each storage cluster node in the network cluster will send a PathRecord request to the subnet manager to query the path information.
[0078] Taking two IB subnets as an example, during ICMP detection, if 2000+ computing nodes send ICMP detection packets to 200 storage nodes, each link (each computing node corresponds to 2 IP addresses stored, a total of 2 links) once per minute, the detection frequency is about 2000*200*2, about 13,000 times per second. For each ICMP detection, in the event of a network failure, the path query will be retried 10 times, then the path query is about 130K / s. If it is a larger-scale storage cluster, more nodes will be affected, and the instantaneous request volume will be even greater. This high-frequency ICMP message detection increases the CPU load of the subnet manager, which may cause the subnet manager process to overload or freeze, thereby paralyzing the entire network system.
[0079] The existing IB subnet management technology cannot automatically control the load of the subnet manager, and the subnet manager itself does not have a mechanism to limit this high load. When the above scenario occurs, the entire network system is paralyzed, and the storage system's external services are interrupted, which will seriously affect the stability of the storage system. The current recovery method is to manually reduce the instantaneous load of the subnet manager by restarting the computing nodes in batches, so that the subnet manager can provide services normally and ensure that the cluster network returns to normal. The operating efficiency of the subnet manager of the network cluster is low. Therefore, how to improve the operating efficiency of the subnet manager of the network cluster is of great practical significance.
[0080] To solve the existing technical problems, the embodiments of the present application provide a load control method, device, medium and electronic device for a subnet manager, which are applied to a storage cluster node of a network cluster based on an infinite bandwidth network; the network cluster includes a subnet manager, a storage cluster node and a computing node; the method includes: generating load record information of the subnet manager according to CPU load information; the load record information includes the received CPU load information of the subnet manager; the CPU load information includes the first CPU load information periodically sent by the subnet manager to the storage cluster node; determining the target operation mode of the subnet manager according to the load record information and a preset load threshold; and sending detection management information to the computing node according to the target operation mode, so that the computing node limits the detection operation on the subnet manager according to the detection management information. The method generates load record information by storing the CPU load information of the subnet manager received by the cluster node; determines the target operation mode of the subnet manager according to the load record information and a preset load threshold; sends detection management information to the computing node according to the target operation mode, so that the computing node limits the detection operation for the subnet manager according to the detection management information, and provides a load control mechanism for the subnet manager of a network cluster, which can efficiently adjust the CPU load of the subnet manager, reduce the high load of the CPU of the subnet manager, and improve the operation efficiency of the subnet manager of the network cluster.
[0081] In order to make the invention purpose, technical scheme and advantages of the embodiments of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present application.
[0082] The technical solution provided in the embodiments of the present application will be described in detail below with reference to the accompanying drawings.
[0083] The load control method of the subnet manager provided in the embodiment of the present application can be applied to a server or a terminal device. The following embodiments of the present application are described by taking the load control method of the subnet manager applied to a terminal device as an example.
[0084] The present application embodiment provides a load control method for a subnet manager, which is applied to a storage cluster node of a network cluster based on an infinite bandwidth network; the network cluster includes a subnet manager, a storage cluster node and a computing node; Figure 1 As shown, the following steps are included:
[0085] Step S101: Generate load record information of the subnet manager according to CPU load information.
[0086] The load record information includes the received CPU load information of the subnet manager; the CPU load information is sent by the subnet manager to the storage cluster node.
[0087] In specific implementation, the storage cluster node generates the load record information of the subnet manager according to the CPU load information; the load record information includes the received CPU load information of the subnet manager; the CPU load information is sent by the subnet manager to the storage cluster node.
[0088] In some embodiments of the present application, the CPU load information includes first CPU load information periodically sent by the subnet manager to the storage cluster node.
[0089] In some embodiments of the present application, the Stor SM Agent of all subnet managers periodically reports the CPU load of the subnet manager process to the Stor Master of the storage cluster node.
[0090] In some optional embodiments, the CPU load information further includes: second CPU load information of the subnet manager fed back to the storage cluster node when the subnet manager receives a load query request sent by the storage cluster node.
[0091] In some embodiments of the present application, if the Stor Master of a storage cluster node does not receive a report from the Stor SMAgent for a long time, the Stor Master proactively queries the Stor SM Agent for the CPU load of the opensm process of the node where it is located.
[0092] In some embodiments of the present application, in order to ensure smooth mode switching, during the target operation mode of the subnet manager, the load record information is determined by using a sampling average value combined with a second-order recursive filter.
[0093] A typical second-order recursive filter can be expressed as:
[0094] F i =α×L i +β×F i-1 +γ×F i-2 ,
[0095] in:
[0096] F i is the estimated value after the i-th filtering;
[0097] L i is the load measurement value at the ith time;
[0098] F i-1 and F i-2 are the first two filtering estimation values respectively;
[0099] α, β, and γ are filter coefficients, and their sum is 1. If you need the output to quickly reflect the most recent changes, you can increase α. If you want the output to change smoothly without rapid fluctuations, you can increase β and γ.
[0100] In some embodiments of the present application, the load record information includes CPU load information and filter parameters, wherein the filter parameters are determined by using a sampling average value combined with a second-order recursive filter; and the filter parameters represent the estimated value after filtering.
[0101] In some embodiments of the present application, the sampled average value may be the load measurement value of the aforementioned embodiment.
[0102] In some embodiments, the filter parameters are obtained as follows:
[0103] The average value is calculated for every N sample values, which is used as the L of the second-order filtering algorithm. i ; The value of N can be adjusted by configuring parameters; α determines the weight of this sampling, and β and γ determine the weight of the historical estimation value;
[0104] The estimated value F is calculated based on the previous two estimated values. i , as the filter parameters.
[0105] In some embodiments of the present application, the sampling value is taken from CPU load information.
[0106] Step S102: determining a target operation mode of the subnet manager according to the load record information and a preset load threshold.
[0107] During specific implementation, the storage cluster node determines the target operation mode of the subnet manager according to the load record information and the preset load threshold.
[0108] In some embodiments of the present application, Stor Master decides that the corresponding IB subnet runs different working modes according to the changes in opensm load, and notifies all Stor Client Agent nodes.
[0109] In some embodiments of the present application, a preset working mode is set according to whether to perform a detection operation corresponding to the IB subnet; the working modes include: normal mode and silent mode. In the normal mode, the computing nodes (external clients) in the characterization system can send detection operations corresponding to the IB subnet at any time according to their own needs, such as sending ICMP messages; the silent mode indicates that if the computing node needs to send ICMP messages to the corresponding IB subnet to detect the link status, it needs to send a message to the Stor Master to apply for detection operation authorization.
[0110] In some embodiments of the present application, in silent mode, if the computing node is authorized, the corresponding IB subnet can send a message to apply for detection operation authorization as needed within the timeout period; if the computing node's application for authorization is rejected or the authorization has expired (exceeded the validity period), the corresponding IB subnet cannot send a message to apply for detection operation authorization.
[0111] In some embodiments of the present application, if the computing node's application for authorization is rejected or the authorization has expired (exceeded the validity period), the computing node needs to send a message to the Stor Master again to apply for authorization after a period of time.
[0112] In some embodiments of the present application, the Stor SM Agent of the subnet manager monitors the CPU utilization of the opensm process of the node, and periodically reports the average load and role information over a period of time to the Stor Master of the cluster node; after receiving the load information, the Stor Master records the load information and decides whether to switch the operating mode of the corresponding IB subnet based on the reported load records; if the Stor SM Agent does not report for a long time, the Stor Master actively queries the Stor SM Agent for the CPU load information of the opensm process of the corresponding node.
[0113] In some embodiments, the decision process for determining the target operating mode of the subnet manager includes:
[0114] The estimated value F is calculated based on the previous two estimated values. i If the actual CPU load is significantly different from the actual CPU load, indicating large fluctuations, wait for several reporting cycles before switching the working mode.
[0115] When the number of times the load is continuously higher or lower than the load threshold reaches the control cycle threshold, the storage cluster node determines whether to allow the switching of the working mode; wherein the control cycle threshold is determined based on the reporting cycle and the expected system fault recovery time; the control cycle threshold can be a configuration parameter.
[0116] In some embodiments, when determining whether to allow switching of the working mode, if the load is higher than a threshold, the Stor Master determines to set the corresponding IB subnet to silent mode; if the load is lower than the threshold, the Stor Master determines to set the corresponding IB subnet to normal mode.
[0117] In some optional embodiments, the target operation mode includes a normal mode and a silent mode; in the above step S102, the process of determining the target operation mode of the subnet manager according to the load record information and the preset load threshold is as follows: Figure 2 As shown, this can be achieved by following the steps below:
[0118] Step S201: judging whether the CPU load information of the control cycle threshold as of the current moment is less than the load threshold according to the load record information and the preset load threshold.
[0119] Step S202: If it is monitored that the CPU load information of the control cycle threshold as of the current moment is all less than the load threshold, the target operation mode of the subnet manager is the normal mode.
[0120] Step S203: If it is monitored that the CPU load information of the control cycle threshold as of the current moment is greater than the load threshold, the target operation mode of the subnet manager is the silent mode.
[0121] The method of this embodiment, the target operation mode, includes normal mode and silent mode; according to the load record information and the preset load threshold, the target operation mode of the subnet manager is determined, including: according to the load record information and the preset load threshold, judging whether the control cycle threshold CPU load information as of the current moment is less than the load threshold; if it is monitored that the control cycle threshold CPU load information as of the current moment is less than the load threshold, then the target operation mode of the subnet manager is normal mode; if it is monitored that the control cycle threshold CPU load information as of the current moment is greater than the load threshold, then the target operation mode of the subnet manager is silent mode, and the target operation mode can be accurately set to normal mode and silent mode, the normal mode characterization allows detection operations on the subnet manager, and the silent mode characterization does not allow detection operations on the subnet manager, so as to further achieve more accurate adjustment of the CPU load of the subnet manager based on the normal mode and the silent mode, reduce the high load of the CPU of the subnet manager, and improve the operation efficiency of the subnet manager of the network cluster.
[0122] Step S103: sending detection management information to the computing node according to the target operation mode, so that the computing node limits the detection operation for the subnet manager according to the detection management information.
[0123] In some embodiments of the present application, when it is determined that the switching of the working mode is allowed, if the load is higher than the threshold, the Stor Master notifies all computing nodes corresponding to the IB subnet to enter silent mode by sending detection management information.
[0124] For example, if the load of either Opensm Master or Opensm Slave exceeds the threshold, Stor Master notifies all computing nodes corresponding to the IB subnet to enter silent mode.
[0125] After receiving the message, the Stor Client Agent calls the tool to shut down the detection of the IB subnet corresponding to this node. If the notification fails, the Stor Master will retry multiple times. After the shutdown is successful, it will record the IB subnet detection status of each computing node.
[0126] In some optional embodiments, according to the target operation mode, sending the detection management information to the computing node so that the computing node limits the detection operation for the subnet manager according to the detection management information includes:
[0127] If the target operation mode is the normal mode, sending the first detection management information to the computing node, so that the computing node is allowed to perform the detection operation for the subnet manager according to the first detection management information;
[0128] If the target operation mode is the silent mode, the second detection management information is sent to the computing node, so that the computing node prohibits the execution of the detection operation for the subnet manager according to the second detection management information.
[0129] The method of this embodiment can further achieve more efficient adjustment of the CPU load of the subnet manager based on the normal mode and the silent mode, reduce the high load of the CPU of the subnet manager, and improve the operating efficiency of the subnet manager of the network cluster.
[0130] In some optional embodiments, the load control method of the subnet manager, such as Figure 3 As shown, the following steps are also included:
[0131] Step S301: If the target operation mode is the silent mode and a detection request sent by the first computing node is received, it is determined whether the authorization queue is full.
[0132] In some embodiments of the present application, the Stor Master of the storage cluster node maintains an authorization queue, the length of which can be adjusted by parameters, and upon receiving an application, it first checks whether the authorization queue is full.
[0133] In specific implementation, when the IB subnet corresponding to the first computing node is in silent mode and there is a detection demand, the Stor Client Agent of the first computing node will initiate a detection request for the corresponding IB subnet. When the storage cluster node receives the detection request sent by the first computing node, it determines whether the authorization queue is full.
[0134] Step S302, if the result of the judgment is that the authorization queue is not full, the first computing node is added to the authorization queue, and the authorization control time of the first computing node is set; the authorization control time is used to remove the first computing node from the authorization queue after the authorization control time is reached to terminate the authorization of the first computing node.
[0135] In specific implementation, if the authorization queue is not full, the corresponding computing node will be added to the authorization queue and the authorization time will be recorded. After the Stor Client Agent obtains the authorization, it will start the detection operation of the corresponding IB subnet. If the corresponding IB subnet connection is restored, the Stor Master will be notified to release the authorization. The Stor Master will remove the node from the authorization queue and record its connection recovery status. If the authorization is not released within the timeout period (which can be adjusted by parameters), the Stor Master will also remove the node from the authorization queue, and the Stor Client Agent will automatically close the detection operation of the corresponding computing node IB subnet after the timeout period.
[0136] Step S303, if the result of the judgment is that the authorization queue is full, the first computing node is refused to be added to the authorization queue, and the authorization blocking time of the first computing node is set; the authorization blocking time is used to prohibit the first computing node from sending a detection request again before the authorization organization time is reached.
[0137] In specific implementation, if the authorization queue is full, authorization will be rejected and the Stor Client Agent will wait for a period of time t before reapplying for authorization; t increases randomly with the number of failed authorization applications to prevent frequent requests to the Stor Master from causing high load on the Stor Master.
[0138] In some embodiments of the present application, it also includes: the storage cluster node records the recovery status of the IB subnets corresponding to all computing nodes; when the recorded recovery status of the IB subnets corresponding to all computing nodes is recovery, and the load is lower than the threshold for several consecutive times, the storage cluster node batch notifies the IB subnets corresponding to all computing nodes to recover from silent mode to normal mode; after receiving the request, the computing node will turn on the detection operation of the IB subnet corresponding to the node and return the result. After all computing nodes respond, it indicates that the corresponding IB subnet has successfully recovered to normal mode; if during the recovery process of the IB subnet from silent mode to normal mode, it is monitored that the load exceeds the threshold again, the target operation mode of the subnet manager is re-determined to be silent mode.
[0139] Exemplarily, when the recovery status of the IB subnets corresponding to all computing nodes recorded by Stor Master is recovery, and the load is lower than the threshold for several consecutive times (wherein the loads of Opensm Master and Opensm Slave are both lower than the threshold), Stor Master will batch notify all Stor Client Agents to restore the corresponding IB subnets from silent mode to normal mode. After receiving the request, Stor Client Agent will turn on the detection operation of the IB subnet corresponding to this node and return the result. After all computing nodes respond, it means that the corresponding IB subnet has successfully recovered to normal mode. If the load exceeds the threshold again during the recovery process, it will re-enter silent mode.
[0140] In some embodiments of the present application, the detection operation is specifically ICMP detection.
[0141] In the method of this embodiment, a control mechanism for computing nodes to initiate detection requests in silent mode is provided, which can further achieve more fine-grained adjustment of the CPU load of the subnet manager based on normal mode and silent mode, reduce the high load of the CPU of the subnet manager, while taking into account the performance stability of the network cluster, and improving the operating efficiency of the subnet manager of the network cluster.
[0142] The present application embodiment provides a load control method for a subnet manager, which is applied to a subnet manager of a network cluster based on an infinite bandwidth network; the network cluster includes a subnet manager, a storage cluster node and a computing node; Figure 4 As shown, the following steps are included:
[0143] Step S401, monitoring the CPU load information of the subnet manager.
[0144] During specific implementation, the subnet manager monitors the CPU load information of the subnet manager.
[0145] Step S402, sending the CPU load information of the subnet manager to the storage cluster node, so that the storage cluster node generates the load record information of the subnet manager according to the CPU load information, and determines the target operation mode of the subnet manager according to the load record information of the subnet manager and a preset load threshold, and sends detection management information to the computing node according to the target operation mode; the detection management information is used for the computing node to limit the detection operation on the subnet manager; the load record information includes the received CPU load information of the subnet manager.
[0146] During specific implementation, the subnet manager sends the CPU load information of the subnet manager to the storage cluster node, so that the storage cluster node generates the load record information of the subnet manager based on the CPU load information, and determines the target operation mode of the subnet manager based on the load record information of the subnet manager and a preset load threshold, and sends detection management information to the computing node according to the target operation mode; the detection management information is used by the computing node to limit the detection operation on the subnet manager; the load record information includes the received CPU load information of the subnet manager.
[0147] In some optional embodiments, the CPU load information includes first CPU load information periodically sent by the subnet manager to the storage cluster node.
[0148] In some optional embodiments, the CPU load information further includes second CPU load information; the second CPU load information is fed back to the storage cluster node by the subnet manager in response to receiving a load query request sent by the storage cluster node.
[0149] The method of this embodiment monitors the CPU load information of the subnet manager through the subnet manager; sends the CPU load information of the subnet manager to the storage cluster node, so that the storage cluster node generates the load record information of the subnet manager according to the CPU load information, and determines the target operation mode of the subnet manager according to the load record information of the subnet manager and a preset load threshold, and sends detection management information to the computing node according to the target operation mode; the detection management information is used for the computing node to limit the detection operation on the subnet manager, and provides a load control mechanism for the subnet manager of the network cluster, which can efficiently adjust the CPU load of the subnet manager, reduce the high load of the CPU of the subnet manager, and improve the operation efficiency of the subnet manager of the network cluster.
[0150] The present application embodiment provides a load control method for a subnet manager, which is applied to a computing node of a network cluster based on an infinite bandwidth network; the network cluster includes a subnet manager, a storage cluster node and a computing node; Figure 5 As shown, the following steps are included:
[0151] Step S501, monitor the detection management information sent by the storage cluster node; the detection management information is generated by the storage cluster node according to the target operation mode and sent to the computing node; the target operation mode is determined according to the load record information and the preset load threshold; the load record information is generated according to the CPU load information; the load record information includes the CPU load information of the subnet manager received by the storage cluster node; the CPU load information is sent by the subnet manager to the storage cluster node.
[0152] Step S502: If the detection management information is received, the detection operation on the subnet manager is restricted according to the received detection management information.
[0153] In some embodiments of the present application, the target operation mode includes a normal mode and a silent mode; the detection management information includes a first detection management information and a second detection management information; in step S502, according to the received detection management information, the process of limiting the detection operation for the subnet manager, such as Figure 6 As shown, this can be achieved through the following process:
[0154] Step S601: If the detection management information is the first detection management information, the detection operation for the subnet manager is allowed to be performed; the first detection management information is sent by the storage cluster node to the computing node when the target operation mode is the normal mode.
[0155] Step S602: If the detection management information is the second detection management information, the detection operation on the subnet manager is prohibited; the second detection management information is sent by the storage cluster node to the computing node when the target operation mode is the silent mode.
[0156] The method of this embodiment, by setting the target operating mode to normal mode and silent mode, the normal mode characterization allows detection operations on the subnet manager, and the silent mode characterization does not allow detection operations on the subnet manager, thereby being able to further achieve more accurate adjustment of the CPU load of the subnet manager based on the normal mode and the silent mode, reduce the high load of the CPU of the subnet manager, and improve the operating efficiency of the subnet manager of the network cluster.
[0157] The load control method of the subnet manager provided in the embodiment of the present application monitors the detection management information sent by the storage cluster node through the computing node; if the detection management information is received, the detection operation on the subnet manager is limited according to the received detection management information; wherein the detection management information is generated by the storage cluster node according to the target operation mode and sent to the computing node; the target operation mode is determined according to the load record information and the preset load threshold; the load record information is generated according to the CPU load information; the load record information includes the CPU load information of the subnet manager received by the storage cluster node; the CPU load information includes the first CPU load information periodically sent by the subnet manager to the storage cluster node, and provides a load control mechanism for the subnet manager of the network cluster, which can efficiently adjust the CPU load of the subnet manager, reduce the high load of the CPU of the subnet manager, and improve the operation efficiency of the subnet manager of the network cluster.
[0158] Based on the same inventive concept, a load control device for a subnet manager is also provided in the embodiment of the present application. Since the device corresponds to the load control method for a subnet manager provided in the embodiment of the present application, and the principle of solving the problem by the device is similar to that of the method, the implementation of the device can refer to the implementation of the above method, and the repeated parts will not be repeated.
[0159] Figure 7 A schematic diagram of the structure of a load control device of a subnet manager provided in an embodiment of the present application is shown. The load control device of the subnet manager is applied to a storage cluster node of a network cluster based on an infinite bandwidth network; the network cluster includes a subnet manager, a storage cluster node and a computing node; Figure 7 As shown, the device includes an information management unit 701, a mode determination unit 702 and a load control unit 703.
[0160] The information management unit 701 is used to generate the load record information of the subnet manager according to the CPU load information; the load record information includes the received CPU load information of the subnet manager; the CPU load information is sent by the subnet manager to the storage cluster node;
[0161] A mode determination unit 702, configured to determine a target operation mode of the subnet manager according to load record information and a preset load threshold;
[0162] The load control unit 703 is used to send detection management information to the computing node according to the target operation mode, so that the computing node limits the detection operation for the subnet manager according to the detection management information.
[0163] In an optional embodiment, the target operation mode includes a normal mode and a silent mode; the mode determination unit 702 is specifically configured to:
[0164] According to the load record information and the preset load threshold, it is determined whether the CPU load information of the control cycle threshold as of the current moment is less than the load threshold;
[0165] If the CPU load information of the control cycle threshold as of the current moment is monitored to be less than the load threshold, the target operation mode of the subnet manager is the normal mode;
[0166] If it is monitored that the control cycle threshold CPU load information as of the current moment is greater than the load threshold, the target operation mode of the subnet manager is the silent mode.
[0167] In an optional embodiment, if Figure 8 As shown, the device further includes an authorization control unit 801; the authorization control unit 801 is used to:
[0168] If the target operation mode is the silent mode and a probe request sent by the first computing node is received, determining whether the authorization queue is full;
[0169] If the result of the judgment is that the authorization queue is not full, the first computing node is added to the authorization queue, and the authorization control time of the first computing node is set; the authorization control time is used to remove the first computing node from the authorization queue after the authorization control time is reached, so as to terminate the authorization of the first computing node;
[0170] If the result of the judgment is that the authorization queue is full, the first computing node is refused to be added to the authorization queue, and the authorization blocking time of the first computing node is set; the authorization blocking time is used to prohibit the first computing node from sending a detection request again before the authorization organization time is reached.
[0171] Fig. 9 A schematic diagram of the structure of a load control device of a subnet manager provided in an embodiment of the present application is shown. The load control device of the subnet manager is applied to a subnet manager of a network cluster based on an infinite bandwidth network; the network cluster includes a subnet manager, a storage cluster node and a computing node; Fig. 9 As shown, the device includes an information monitoring unit 901 and an information reporting unit 902.
[0172] Wherein, the information monitoring unit is used to monitor the CPU load information of the subnet manager;
[0173] An information reporting unit is used to send the CPU load information of the subnet manager to the storage cluster node, so that the storage cluster node generates the load record information of the subnet manager according to the CPU load information, and the storage cluster node determines the target operation mode of the subnet manager according to the load record information of the subnet manager and a preset load threshold, and sends detection management information to the computing node according to the target operation mode; the detection management information is used for the computing node to limit the detection operation on the subnet manager; the load record information includes the received CPU load information of the subnet manager.
[0174] Fig.10 A schematic diagram of the structure of a load control device of a subnet manager provided in an embodiment of the present application is shown. The load control device of the subnet manager is applied to a computing node of a network cluster based on an infinite bandwidth network; the network cluster includes a subnet manager, a storage cluster node and a computing node; Fig.10 As shown, the device includes a management information monitoring unit 1001 and a received information response unit 1002.
[0175] Wherein, the management information monitoring unit 1001 is used to monitor the detection management information sent by the storage cluster node; the detection management information is generated by the storage cluster node according to the target operation mode and sent to the computing node; the target operation mode is determined according to the load record information and the preset load threshold; the load record information is generated according to the CPU load information; the load record information includes the CPU load information of the subnet manager received by the storage cluster node; the CPU load information is sent by the subnet manager to the storage cluster node;
[0176] The received information response unit 1002 is used to limit the detection operation for the subnet manager according to the received detection management information if the detection management information is received.
[0177] Based on the same inventive concept, the present application also provides a load control system for a subnet manager. Since the system includes the subnet manager, storage cluster node and computing node provided by the method of the aforementioned embodiment of the present application, and the principle of solving the problem by the system is similar to that of the method, the implementation of the system can refer to the implementation of the aforementioned method, and the repeated parts will not be repeated.
[0178] Based on the same inventive concept as the above method embodiment, an electronic device is also provided in the embodiment of the present application. The electronic device can be used for load control of a subnet manager. In the embodiment of the present application, the electronic device can be a server or a terminal device. In one embodiment, the electronic device can be a terminal device. In this embodiment, the structure of the electronic device can be as follows: Fig.11 As shown, it includes a memory 1101 , a communication module 1103 and one or more processors 1102 .
[0179] The memory 1101 is used to store computer programs executed by the processor 1102. The memory 1101 may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system and programs required for running the instant messaging function, etc.; the data storage area may store various instant messaging information and operation instruction sets, etc.
[0180] The memory 1101 may be a volatile memory, such as a random-access memory (RAM); the memory 1101 may also be a non-volatile memory, such as a read-only memory, a flash memory, a hard disk drive (HDD) or a solid-state drive (SSD), or the memory 1101 is any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto. The memory 1101 may be a combination of the above memories.
[0181] The processor 1102 may include one or more central processing units (CPU) or a digital processing unit, etc. The processor 1102 is configured to implement the above-mentioned load control method of the subnet manager when calling the computer program stored in the memory 1101 .
[0182] The communication module 1103 is used to communicate with terminal devices and other servers.
[0183] The specific connection medium between the memory 1101, the communication module 1103 and the processor 1102 is not limited in the embodiment of the present application. Fig.11 In the embodiment, the memory 1101 and the processor 1102 are connected via a bus 1104. The bus 1104 is connected to the processor 1102 via a bus 1104. Fig.11 The connections between other components are shown in bold lines, which are only for illustration and are not intended to be limiting. The bus 1104 can be divided into an address bus, a data bus, a control bus, etc. For ease of illustration, Fig.11 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.
[0184] According to one aspect of the present application, a computer program product or computer program is provided, which includes computer instructions, which are stored in a computer-readable storage medium. The processor of the computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the load control method of the subnet manager in the above embodiment. The program product can adopt any combination of one or more readable media. The readable medium can be a readable signal medium or a readable storage medium. The readable storage medium can be, for example, - but not limited to - an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of readable storage media (a non-exhaustive list) include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.
[0185] The above description is only a specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application.
Claims
1. A load control method for a subnet manager, characterized in that: A storage cluster node applied to a network cluster based on an infinite bandwidth network; the network cluster includes a subnet manager, the storage cluster node and a computing node; the method includes: Generate load record information of the subnet manager according to the CPU load information; the load record information includes the received CPU load information of the subnet manager; the CPU load information is sent by the subnet manager to the storage cluster node; Determining a target operation mode of the subnet manager according to the load record information and a preset load threshold; According to the target operation mode, detection management information is sent to the computing node, so that the computing node limits the detection operation for the subnet manager according to the detection management information.
2. The method according to claim 1, characterized in that The target operation mode includes a normal mode and a silent mode; the determining the target operation mode of the subnet manager according to the load record information and a preset load threshold includes: According to the load record information and the preset load threshold, determining whether the control cycle threshold CPU load information up to the current moment is all less than the load threshold; If it is monitored that the CPU load information of the control cycle threshold as of the current moment is all less than the load threshold, the target operation mode of the subnet manager is the normal mode; If it is monitored that the control cycle threshold CPU load information up to the current moment is greater than the load threshold, the target operation mode of the subnet manager is the silent mode.
3. The method according to claim 2, characterized in that The method further comprises: If the target operation mode is the silent mode and a detection request sent by the first computing node is received, determining whether the authorization queue is full; If the result of the judgment is that the authorization queue is not full, the first computing node is added to the authorization queue, and an authorization control time of the first computing node is set; the authorization control time is used to remove the first computing node from the authorization queue after the authorization control time is reached, so as to terminate the authorization of the first computing node; If the result of the judgment is that the authorization queue is full, the first computing node is refused to be added to the authorization queue, and the authorization blocking time of the first computing node is set; the authorization blocking time is used to prohibit the first computing node from sending a detection request again before the authorization organization time is reached.
4. A load control method for a subnet manager, characterized in that: A subnet manager applied to a network cluster based on an infinite bandwidth network; the network cluster includes the subnet manager, a storage cluster node and a computing node; the method includes: Monitoring the CPU load information of the subnet manager; Sending the CPU load information of the subnet manager to the storage cluster node, so that the storage cluster node generates the load record information of the subnet manager according to the CPU load information, and determines the target operation mode of the subnet manager according to the load record information of the subnet manager and a preset load threshold, and sending detection management information to the computing node according to the target operation mode; the detection management information is used by the computing node to limit the detection operation on the subnet manager; the load record information includes the received CPU load information of the subnet manager.
5. A load control method for a subnet manager, characterized in that: The method is applied to a computing node of a network cluster based on an infinite bandwidth network; the network cluster includes a subnet manager, a storage cluster node and the computing node; the method includes: Monitoring the detection management information sent by the storage cluster node; The detection management information is generated by the storage cluster node according to the target operation mode and sent to the computing node; the target operation mode is determined according to the load record information and a preset load threshold; the load record information is generated according to the CPU load information; the load record information includes the CPU load information of the subnet manager received by the storage cluster node; the CPU load information is sent by the subnet manager to the storage cluster node; If the detection management information is received, the detection operation on the subnet manager is restricted according to the received detection management information.
6. The method according to claim 5, characterized in that The target operation mode includes a normal mode and a silent mode; the detection management information includes first detection management information and second detection management information; The limiting the detection operation for the subnet manager according to the received detection management information includes: If the detection management information is the first detection management information, the detection operation for the subnet manager is allowed to be performed; the first detection management information is sent by the storage cluster node to the computing node when the target operation mode is the normal mode; If the detection management information is the second detection management information, the detection operation on the subnet manager is prohibited; the second detection management information is sent by the storage cluster node to the computing node when the target operation mode is the silent mode.
7. A load control device for a subnet manager, characterized in that: A storage cluster node applied to a network cluster based on an infinite bandwidth network; the network cluster includes a subnet manager, the storage cluster node and a computing node; the device includes: An information management unit, configured to generate load record information of the subnet manager according to the CPU load information; the load record information includes the received CPU load information of the subnet manager; the CPU load information is sent by the subnet manager to the storage cluster node; A mode determination unit, configured to determine a target operation mode of the subnet manager according to the load record information and a preset load threshold; A load control unit is used to send detection management information to the computing node according to the target operation mode, so that the computing node limits the detection operation for the subnet manager according to the detection management information.
8. A load control device for a subnet manager, characterized in that: A subnet manager applied to a network cluster based on an infinite bandwidth network; the network cluster includes the subnet manager, a storage cluster node and a computing node; the device includes: An information monitoring unit, used to monitor the CPU load information of the subnet manager; An information reporting unit is used to send the CPU load information of the subnet manager to the storage cluster node, so that the storage cluster node generates the load record information of the subnet manager according to the CPU load information, and the storage cluster node determines the target operation mode of the subnet manager according to the load record information of the subnet manager and a preset load threshold, and sends detection management information to the computing node according to the target operation mode; the detection management information is used by the computing node to limit the detection operation on the subnet manager; the load record information includes the received CPU load information of the subnet manager.
9. A load control device for a subnet manager, characterized in that: A computing node applied to a network cluster based on an infinite bandwidth network; the network cluster includes a subnet manager, a storage cluster node and the computing node; the device includes: A management information monitoring unit, used to monitor the detection management information sent by the storage cluster node; The detection management information is generated by the storage cluster node according to the target operation mode and sent to the computing node; the target operation mode is determined according to the load record information and a preset load threshold; the load record information is generated according to the CPU load information; the load record information includes the CPU load information of the subnet manager received by the storage cluster node; the CPU load information is sent by the subnet manager to the storage cluster node; The receiving information response unit is used to limit the detection operation for the subnet manager according to the received detection management information if the detection management information is received.
10. A computer-readable storage medium, wherein a computer program is stored in the computer-readable storage medium, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 3, the method according to claim 4, or the method according to any one of claims 5 to 6 is implemented.
11. An electronic device, characterized in that: The method comprises a memory and a processor, wherein the memory stores a computer program executable on the processor, and when the computer program is executed by the processor, the method according to any one of claims 1 to 3, the method according to claim 4, or the method according to any one of claims 5 to 6 is implemented.