WO2022063273A1

WO2022063273A1 - Resource allocation method and apparatus based on numa attribute

Info

Publication number: WO2022063273A1
Application number: PCT/CN2021/120706
Authority: WO
Inventors: 贾金科; 徐安; 秦洲
Original assignee: 华为云计算技术有限公司
Priority date: 2020-09-27
Filing date: 2021-09-26
Publication date: 2022-03-31
Also published as: CN114281516A

Abstract

A resource allocation method and apparatus based on a NUMA attribute, which are applied to the field of cloud computing. The method comprises: obtaining resource allocation policies issued by a cloud management platform and NUMA attributes of management resources; and allocating central processing units (CPUs) of a computing node to the management resources according to the resource allocation policies and the NUMA attributes of the management resources. By means of the method of the present application, a computing node can configure different resource allocation policies according to NUMA attributes of management resources, thereby improving the flexibility of resource allocation while also improving the resource utilization rate.

Description

A NUMA attribute-based resource allocation method and device

technical field

The present application relates to the field of cloud computing technologies, and in particular, to a method and device for resource allocation based on NUMA attributes.

Background technique

At present, the server architecture used by the cloud management platform is a non-uniform memory access (NUMA) architecture. Based on the NUMA architecture, the central processing unit (CPU) and memory of the server can be divided into multiple NUMA nodes. Among them, each NUMA node has multiple CPU cores, some CPU cores are used for running virtual machines, and the other CPU cores are used for running management processes of the cloud management platform.

Since the number of CPU cores occupied by the management process on different NUMA nodes is different, the available resources of the virtual machine are uncertain, which will lead to excessive resource fragmentation when the virtual machine is issued, resulting in low resource utilization.

SUMMARY OF THE INVENTION

The present application provides a NUMA attribute-based resource allocation method and device to improve resource utilization.

In a first aspect, the present application provides a resource allocation method based on NUMA attributes, which can be applied to computing nodes. Specifically, the method includes: obtaining a resource allocation policy issued by a cloud management platform and NUMA attributes of management resources; The acquired resource allocation policy and the NUMA attribute of the management resource allocate the central processing unit CPU of the computing node for the management resource.

Through the above technical solution, the computing node can configure different resource allocation strategies according to the NUMA attributes of the management resources, which can improve the flexibility of resource allocation, reduce resource fragmentation, and improve resource utilization.

In a possible design, the management resources include a first management resource, a second management resource, and a third management resource. The attributes of the first management resource include: the CPU of the first NUMA node of the computing node needs to be allocated to the first management resource; the attributes of the second management resource include: the CPU of any NUMA node in the computing nodes can be allocated to the second management resource. Management resource; the attributes of the third management resource include: CPUs can be allocated to the third management resource across NUMA nodes in the computing node.

In the above technical solution, different management resources have different NUMA attributes, and for different management resources, the assignable NUMA nodes may be the same or different, so that management resources can be allocated to NUMA nodes according to the NUMA node attributes of the management resources.

In a possible design, the resource allocation strategy includes a balanced resource allocation strategy, and according to the resource allocation strategy and the NUMA attribute of the management resource, the central processing unit CPU of the computing node is allocated to the management resource, including :

Allocate the first management resource, the second management resource, and the third management resource according to the NUMA attributes corresponding to the first management resource, the second management resource, and the third management resource, respectively. to the first NUMA node, the second NUMA node, and the third NUMA node.

Through the above technical solution, the computing node can allocate the management resources to different NUMA nodes according to the NUMA attributes of the management resources, and when the management resources are allocated according to the balanced resource allocation strategy, the management resources on different NUMA nodes can be occupied by the management resources. The CPU cores are balanced, which can reduce resource fragmentation on computing nodes and improve resource utilization. In addition, the CPU cores available to the virtual machine on each NUMA node can be equalized, thereby improving the creation specification of the virtual machine.

In a possible design, the method further includes: after the third management resource is allocated to the third NUMA node, if the third management resource still needs to occupy H CPU cores, the first The NUMA node and the second NUMA node, whichever has the most remaining CPU cores, allocates H CPU cores to the third management resource, or allocates H/2 CPU cores from the first NUMA node and the second NUMA node to the third management resource. management resources, or allocating H CPU cores from the fourth NUMA node to the third management resources.

Through the above technical solution, the computing node can allocate the remaining third management resources to NUMA nodes with relatively few CPU cores according to the number of CPU cores occupied by the remaining third management resources, or can also allocate the remaining third management resources to NUMA nodes that occupy less CPU cores. Resources are evenly distributed to other NUMA nodes, which can ensure that the CPU cores occupied by management resources on different NUMA nodes are balanced and improve resource utilization.

In a possible design, the resource allocation strategy includes a centralized resource allocation strategy, and according to the centralized resource allocation strategy and the NUMA attribute of the management resource, the CPU of the computing node is allocated to the management resource, The method includes: sequentially classifying the first management resource, the second management resource and the third management resource according to the NUMA attributes corresponding to the first management resource, the second management resource and the third management resource respectively Resources are allocated to the first NUMA node.

Through the above technical solution, the computing node can centrally allocate management resources of different NUMA attributes to one NUMA node, so that the performance of the computing node can be improved, and a larger size virtual machine can be created on a single NUMA.

In a possible design, the method further includes: after the third management resource is allocated to the first NUMA node, if the third management resource still needs to occupy H CPU cores, the second The NUMA node or the third NUMA node allocates H CPU cores to the third management resource.

Through the above technical solution, if the third management resources are not allocated, the remaining third management resources can be allocated to other NUMA nodes, thereby avoiding resource waste and improving resource utilization.

In a second aspect, the present application provides a resource allocation device based on NUMA attributes, including: an acquisition unit, where the acquisition unit is configured to acquire a resource allocation strategy issued by a cloud management platform and NUMA attributes of management resources; an allocation unit, used for The central processing unit CPU of the computing node is allocated to the management resource according to the resource allocation policy and the NUMA attribute of the management resource obtained by the obtaining unit.

In a possible design, the management resource includes a first management resource, a second management resource, and a third management resource; the attributes of the first management resource include: the first NUMA node to which the computing node needs to be allocated. CPU to the first management resource; attributes of the second management resource include: the CPU of any NUMA node of the computing node can be allocated to the second management resource; attributes of the third management resource include: CPUs may be allocated to the third management resource in the compute nodes across NUMA nodes.

In a possible design, when the resource allocation strategy includes a balanced resource allocation strategy, the allocation unit is specifically configured to, according to the resource allocation strategy and the NUMA attribute of the management resource, provide the The management resource is allocated to the central processing unit CPU of the computing node: according to the NUMA attributes corresponding to the first management resource, the second management resource, and the third management resource, respectively, the first management resource, all the The second management resource and the third management resource are allocated to the first NUMA node, the second NUMA node and the third NUMA node.

In a possible design, the allocating unit is further configured to: after the third management resource is allocated to the third NUMA node, if the third management resource still needs to occupy H CPU cores, then The first NUMA node and the second NUMA node, whichever has the most remaining CPU cores, allocate H CPU cores to the third management resource, or allocate H/2 CPU cores from the first NUMA node and the second NUMA node to the third management resource. The third management resource, or H CPU cores are allocated from the fourth NUMA node to the third management resource.

In a possible design, when the resource allocation strategy includes a centralized resource allocation strategy, the allocation unit is specifically configured to, according to the centralized resource allocation strategy and the NUMA attribute of the management resource, be as follows: The management resource is allocated to the CPU of the computing node: according to the NUMA attributes corresponding to the first management resource, the second management resource and the third management resource, the first management resource, the The second management resource and the third management resource are allocated to the first NUMA node.

In a possible design, the allocating unit is further configured to: after the third management resource is allocated to the first NUMA node, if the third management resource still needs to occupy H CPU cores, then The second NUMA node or the third NUMA node allocates H CPU cores to the third management resource.

In a third aspect, the present application provides a NUMA attribute-based resource allocation apparatus, which has the function of implementing the resource allocation method in the first aspect or any possible implementation manner of the first aspect. The functions can be implemented by hardware, or can be implemented by hardware executing corresponding software.

The apparatus includes a communication interface for receiving and sending data, a processor and a memory, the processor being configured to support the apparatus to perform the first aspect or any of the possible implementations of the first aspect corresponding function. The memory is coupled to the processor and holds program instructions necessary for the apparatus.

In a fourth aspect, a computer-readable storage medium is provided, where instructions are stored in the computer-readable storage medium, when the computer-readable storage medium runs on a computer, the computer causes the computer to execute the methods in the first aspect and the various embodiments.

In a fifth aspect, there is provided a computer program product comprising instructions, which, when executed on a computer, cause the computer to perform the methods of the first aspect and the various embodiments above.

In a sixth aspect, a chip is provided, and logic in the chip is used to execute the methods in the first aspect and each of the above embodiments.

Regarding the technical effects brought about by the various embodiments of the second aspect to the sixth aspect or the second aspect to the sixth aspect, reference may be made to the introduction to the technical effects of the first aspect or the various embodiments of the first aspect. I won't go into details here.

Description of drawings

1 is a schematic diagram of a system architecture in which an embodiment of the present application is applied to a cloud scenario;

2 is a flowchart of a method for resource allocation based on NUMA attributes provided by an embodiment of the present application;

3 is a flowchart of a method for resource allocation based on NUMA attributes provided by an embodiment of the present application;

FIG. 4 is a schematic diagram of a resource allocation method according to an embodiment of the present application;

5 is a flowchart of a method for resource allocation based on NUMA attributes provided by an embodiment of the present application;

6 is a schematic diagram of a NUMA attribute-based resource allocation apparatus provided by an embodiment of the present application;

FIG. 7 is a schematic diagram of another apparatus for resource allocation based on NUMA attributes provided by an embodiment of the present application.

detailed description

The embodiments of the present application will be described in further detail below with reference to the accompanying drawings.

For ease of understanding, the descriptions of concepts related to the present application are exemplarily given for reference, as follows:

1) Non-uniform memory access NUMA: It is a memory architecture designed for multi-processor computers, and the memory access time depends on the location of the memory relative to the processor. Under NUMA, a processor can access its own local memory a bit faster than non-local memory (memory located on another processor, or memory shared between processors).

2) Virtual machine: Use virtualization technology to simulate a physical host as multiple logical hosts (called virtual machines). A virtual machine refers to a logical host in a complete computer system that is simulated by software with complete hardware system functions and can run in a completely isolated environment.

3) Computing node: In this application, it can be understood as a host or a server.

At least one involved in the embodiments of the present application includes one or more; wherein, multiple refers to greater than or equal to two. In addition, in the description of this application, words such as "first" and "second" are only used for the purpose of distinguishing the description objects, and cannot be understood as indicating or implying relative importance, nor can they be understood as indicating or implying order.

Referring to FIG. 1 , it is a schematic diagram of a system architecture in which an embodiment of the present application is applied to a cloud scenario. Under the system architecture shown in FIG. 1 , the server can be divided into two chips (chips), such as CPU0 and CPU1, wherein each chip can include a NUMA node. Exemplarily, the server shown in FIG. 1 may include 4 NUMA nodes, and each NUMA node has multiple processor cores (core0, core1, etc.). Some of these processor cores can be allocated to management resources and some can be allocated to virtual machines.

Taking NUMA node0 as an example, core0 and core1 can be allocated to management resources, and core2, core3...core31 can be allocated to virtual machines. Of course, it can be understood that the above schematic diagram is only a schematic illustration, and the allocation of processor cores is not limited to this. For example, core0, core1...core20 can also be used for management resources, and core21...core31 can be used by virtual machines, etc. This application does not specifically limit this.

In cloud scenarios, in order to avoid the impact between virtual machine operation and management resources, management resources need to be isolated. Based on the above-mentioned NUMA architecture, when isolation of management resources is performed, it is necessary to isolate different processor cores on different NUMA nodes for use by management resources. Since the number of processor cores allocated to management resources on different NUMA nodes is uncertain, the number of processor cores allocated to virtual machines is uncertain, which will affect the specifications of virtual machines that can be allocated by the server. and quantity, thus affecting the resource utilization of the server.

In view of the above technical problems, in the embodiment of the present application, the system administrator can configure different resource allocation policies for different servers, and the cloud management platform delivers them to different servers. After the server receives the resource allocation policies, the resource allocation policies can be: Manage resources and allocate server cores to virtual machines to maximize the use of server resources and improve resource utilization.

Referring to FIG. 2, a flowchart of a method for resource allocation based on NUMA attributes provided by an embodiment of the present application. Referring to FIG. 2, the method may include the following steps:

S201: Acquire the resource allocation policy issued by the cloud management platform and the NUMA attribute of the management resource.

It should be noted that the management resource may include multiple management processes, and different management processes may correspond to different NUMA attributes.

S202: Allocate the CPU of the computing node for the management resource according to the resource allocation policy and the NUMA attribute of the management resource.

For the convenience of description, the resource allocation strategy may be denoted as "balanced resource allocation strategy" and "centralized resource allocation strategy" in this embodiment of the present application.

It should be understood that the balanced resource allocation strategy can be understood as making the management resources on different NUMA nodes as evenly distributed as possible when resource isolation is performed on management resources, so that different NUMA nodes can provide virtual machines with Resources are equal. The centralized resource allocation strategy can be understood as making the management resources centrally allocated to a certain NUMA node when the management resources are isolated.

In this embodiment of the present application, the management resources may include: a first management resource, a second management resource, and a third management resource. It is assumed that the NUMA nodes on the server (computing node) include: a first NUMA node, a second NUMA node, a third NUMA node, and a fourth NUMA node.

Considering that different management resources are used for resource isolation, the attribute requirements of resource isolation may be different. In this embodiment of the present application, the NUMA attributes of different management resources may correspond to the following:

The attribute of the first management resource includes: the CPU of the first NUMA node of the computing node needs to be allocated to the first management resource; the attribute of the second management resource includes: the CPU of any NUMA node of the computing node can be allocated to the second management resource; The attributes of the third management resource include that CPUs can be allocated to the third management resource across NUMA nodes in the computing node.

Exemplarily, as a possible implementation manner, the above management resources can be divided into the following categories:

The first category: the management resource of a certain NUMA node must be used fixedly, such as resource L.

The second category: a management resource of a NUMA node must be used, such as resource M.

The third category: management resources that can be flexibly allocated on each NUMA node, for example, denoted as resource N.

It can be understood that the first type of resource, namely resource L, may correspond to the above-mentioned first management resource; the second type of resource, namely resource M, may correspond to the above-mentioned second management resource; the third type of resource, namely resource N, may correspond to the above-mentioned third management resource. Manage resources.

In some embodiments of the present application, when a computing node adopts a balanced resource allocation strategy, the CPU of the computing node may be allocated for management resources in the following manner:

According to the NUMA attributes corresponding to the first management resource, the second management resource and the third management resource, respectively, the first management resource, the second management resource and the third management resource are allocated to the first NUMA node, the second NUMA node and the third management resource. Three NUMA nodes. That is to say, the computing node can respectively allocate three types of management resources with different NUMA attributes to different NUMA nodes according to the NUMA attributes.

Considering that the third management resource may occupy more CPU cores, the third management resource may not be fully allocated on the third NUMA node allocated above, and at this time, it is necessary to determine the CPU that the third management resource still needs to occupy The number of cores (for example, H CPU cores need to be occupied).

As a possible implementation manner, if the third management resource still needs to occupy H CPU cores, the first NUMA node and the second NUMA node with the most remaining CPU cores are allocated H CPU cores to the third management resource. That is, the remaining H CPU cores of the third management resource may be allocated to the NUMA node with the most remaining CPU cores among the first NUMA node and the second NUMA node.

As another possible implementation manner, if the third management resource still needs to occupy H CPU cores, H/2 CPU cores may be allocated from the first NUMA node and the second NUMA node to the third management resource. That is, the H CPU cores are equally distributed to the first NUMA node and the second NUMA node.

As another possible implementation manner, if the third management resource still needs to occupy H CPU cores, then H CPU cores may be allocated from the fourth NUMA node to the third management resource. That is, the remaining H CPU cores of the third management resource can be allocated to other NUMA nodes.

The following describes the process of resource allocation using the balanced resource allocation strategy in detail with reference to the accompanying drawings. As shown in FIG. 3 , a flowchart of a method for resource allocation based on NUMA attributes provided by an embodiment of the present application. Referring to FIG. 3 , the method includes the following steps:

S301: Divide the CPU and memory on the server into Q NUMA nodes.

Among them, Q is a positive integer greater than or equal to 1. In this application, when the CPU and memory on the server are divided into Q NUMA nodes, they can be divided in the form of equal division. In other words, the number of CPUs and memory on the server can be evenly divided among Q NUMA nodes, so that the number of CPUs and memory on each NUMA node are equal.

S302: Allocate the resource L to a fixed NUMA node among the Q NUMA nodes.

In the embodiment of the present application, since the resource L is a resource using a fixed NUMA node, it is necessary to first allocate the resource L to the NUMA nodes, and determine which NUMA node the resource L occupies and which NUMA node the resource L occupies in the NUMA node. The number of processor cores (CPUs) accounted for.

It should be noted that the resource L may be a resource that uses a certain NUMA node on a fixed basis, or a resource that uses a certain two NUMA nodes among the Q NUMA nodes, etc., which is not limited in this application.

S303: Allocate the resource M to a NUMA node different from the NUMA node where the resource L is located.

For a resource M that uses one NUMA node fixedly, it can be allocated after fixing the resource L to Q NUMA nodes. Resource M can be allocated to a NUMA node alone, or can be allocated to a NUMA node where resource L is located, as long as it is ensured that resource M is located on the same NUMA node, that is, it is not allocated across NUMA nodes.

As an example, the resource M may be allocated to a different NUMA node from the resource L, so that different NUMNA nodes will have management resources, and there will be no situation where all management resources are allocated to one NUMA node.

S304: Determine the maximum value of the number of CPU cores occupied by the resource L and the resource M.

Assume that the number of CPU cores occupied by resource L is x1, and the number of CPU cores occupied by resource M is x2. Exemplarily, as shown in FIG. 4, for example, resource L is first allocated to Q NUMA nodes, for example, to NUMA0, and then resource M is allocated, for example, resource M can be allocated to NUMA2, then the management resources in NUMA0 occupy the The number of CPU cores is x1, and the number of CPU cores occupied by management resources in NUMA2 is x2.

S305: Allocate the resource N to a NUMA node different from the NUMA node where the resource L and the resource M are located.

Based on the above steps S302 and S303, it can be known that the positions of resource L and resource M in the Q NUMA nodes have been determined. In order to keep the management resources on each NUMA node balanced, the allocation of resource N is given priority to non-resources. L and the NUMA node where non-resource M is located.

In this embodiment of the present application, the resource N may be allocated according to the maximum number of CPU cores occupied by the resource L and the resource M. In other words, the maximum number of CPU cores occupied by resource L and resource M can be taken as the maximum number of CPU cores occupied by management resources set in each NUMA node, and then based on the maximum value. Allocate resource N.

It should be noted that the number of CPU cores occupied by resource L and resource M may also be the same. In this case, resource N may be allocated according to the number of CPU cores occupied by resource L or resource M.

Exemplarily, continuing to refer to FIG. 4 , after the resource L is allocated to NUMA0 and the resource M is allocated to NUMA2, the resource N can be allocated to NUMA3. Of course, the resource N may also be allocated to NUMA1, etc., which is not limited in this application.

As a possible implementation manner, when resource N is allocated, it may be allocated in the order of NUMA nodes. Assume that the maximum number of CPU cores occupied by resource L and resource M is the number of CPU cores occupied by resource L, that is, x1. For example, if resource L is allocated to NUMA0 and resource M is allocated to NUMA2, then resource N can be allocated to NUMA1 first. When the number of CPU cores occupied by resource N in NUMA1 is x1, then resource N can be allocated to NUMA3. . Of course, the resource N may also be randomly allocated to the NUMA nodes that are not resource L and non-resource M, which is not limited in this application.

S306: Determine whether there are remaining unallocated resources N. If there is remaining resource N, continue to step S307.

In this embodiment of the present application, when the number of CPU cores occupied by resource N is relatively small (for example, the number of CPU cores occupied by resource N is less than the set threshold), it can be allocated to NUMA nodes that are not resource L and non-resource M middle. That is to say, resource N can be allocated to NUMA nodes other than resource L and non-resource M, and there will be no remaining unallocated resource N.

When the number of CPU cores occupied by resource N is relatively large (the number of CPU cores occupied by resource N is greater than the set threshold), after allocating in the NUMA nodes of non-resource L and non-resource M, there will be remaining unallocated the resource N.

S307: Calculate the number of CPU cores occupied by management resources in each of the Q NUMA nodes.

When the remaining resource N is allocated in the NUMA nodes that are not resource L and non-resource M, it is necessary to calculate the number of CPU cores occupied by the management resources in each NUMA node. Since there is a surplus of resource N, the number of CPU cores occupied by management resources in the NUMA nodes that are not resource L and non-resource M is the maximum number of CPU cores occupied by resource L and resource M. That is to say, when the resource N cannot be allocated to the NUMA nodes other than the resource L and the resource M, it can be allocated to the NUMA node where the resource L and/or the resource M are located. At this point, the number of CPU cores occupied by the management resources in each NUMA node can be calculated.

Exemplarily, it is assumed that resource L is allocated to NUMA0, resource M is allocated to NUMA2, the number of CPU cores occupied by resource L is 6, and the number of CPU cores occupied by resource M is 4. When the number of CPU cores is 14, in order to ensure that the number of CPU cores occupied by the management resources on each NUMA node is balanced, the resource N can be allocated up to 6 CPU cores on the NUMA1 and NUMA3 that can be allocated, respectively. That is, there are 2 remaining CPU cores left unallocated.

S308: Determine the NUMA node with the least number of CPU cores occupied by management resources among the Q NUMA nodes.

S309: Allocate the remaining resource N to the NUMA node with the smallest number of CPU cores occupied by the management resource.

In order to ensure the balance of the number of CPU cores occupied by management resources on each NUMA node, the remaining resource N can be allocated to the NUMA node with the smallest number of CPU cores occupied by management resources. Exemplarily, assuming that the relationship between the number of CPU cores occupied by resource L and resource M in the NUMA node is: x1>x2, that is to say, the number of CPU cores occupied by management resources in the NUMA2 node is the least, then the The remaining resource N is allocated to NUMA2 nodes.

Continuing to use the above example, if there are 2 CPU cores left unallocated, the 2 CPU cores can be allocated to the NUMA node with more remaining CPU cores among the NUMA nodes, for example, it can be allocated to the NUMA2 node where the resource M is located, so that NUMA0 The CPU cores occupied by management resources on , NUMA1, NUMA2 and NUMA3 are balanced.

Of course, the above allocation method is only an example. In practical applications, the remaining number of CPU cores can also be evenly allocated to NUMA nodes, or a part of them can be allocated to NUMA nodes with more remaining CPU cores, and then The rest is evenly distributed to several NUMA nodes, which is not limited in this application.

Through the above steps, the number of CPU cores occupied by the management resources in each NUMA node can be the same, which is equal to the maximum number of CPU cores occupied by the resource L and the resource M.

S3010: Continue to allocate the CPU cores occupied by the remaining resources N in step S309 to the Q NUMA nodes until the resource N is allocated and the CPUs occupied by the management resources in the Q NUMA nodes are balanced.

If the resource N is still not allocated after step S309 is executed, the number of CPU cores occupied by the remaining resource N in step S308 may be evenly allocated on the Q NUMA nodes. Of course, in the result obtained by dividing the remaining resource N by the number Q of NUMA nodes in step S308, there may be a remainder, that is, it cannot be distributed evenly, then the remainder can be preferentially allocated to the NUMA nodes where the non-resource L and non-resource M are located , if the allocation still cannot be completed, then allocate to the NUMA node where the resource L and/or resource M are located until the resource N is all allocated.

Further, in the embodiment of the present application, in the process of allocating resource N, the number of CPU cores allocated to resource N on each NUMA node may be recorded, and then the memory of the server is allocated to the CPU core occupied by resource N. On the NUMA node with the largest number, to ensure that when accessing memory, there is no need to access across NUMA nodes, shortening the memory access time and improving access efficiency.

The resource allocation method in the embodiment shown in FIG. 3 will be introduced below with reference to specific examples.

Assuming that the host (server) is configured with 96 (pcs) CPUs and 384G of memory, and the CPU and memory are divided into 4 NUMA nodes, such as NUMA0, NUMA1, NUMA2, and NUMA3, the available resources of each NUMA node are 24CPU and 96G of memory.

Exemplarily, the management processes that can be included in this application and the number of CPU cores and memory occupied by each process are as follows:

Virtualization management process: 2CPU, 12G memory; computing management process: 6CPU, 5G memory; network management process: 4CPU, 8G memory; storage management process: 4CPU, 7G memory.

Among them, the memory of the virtualization management process can be evenly distributed among the NUMA nodes; the CPU cores of the computing management process can be dynamically allocated, and the memory needs to be assigned a NUMA node; the network management process must be fixed on the NUMA0 node; the storage management process must be on a NUMA node middle.

It should be understood that the network management process may correspond to the resource L (ie the first management resource), the storage management process may correspond to the resource M (ie the second management resource), and the virtualization management process and the computing management process may correspond to the above-mentioned Resource N (ie the third management resource).

In order to clearly reflect the effect brought by the allocation strategy of the present application, Table 1 is used as an example below to illustrate the allocation of management resources by NUMA nodes on the server before applying the allocation strategy of the present application, as shown in Table 1.

Table 1

It can be understood that the above table is only a schematic illustration, and in practical applications, the allocation form of management resources is not limited to the above examples.

It can be seen from Table 1 that the management resources on NUMA0 occupy the most CPU cores, and the maximum size of virtual machines that can be created by this server is 4 NUMA0 virtual machines, that is, 75CPU and 312G memory.

After utilizing the balanced resource allocation strategy in this application, the allocation of management resources by NUMA nodes on the server can be referred to as shown in Table 2.

Table 2

It can be seen from Table 2 that after using the balanced resource allocation strategy, the number of CPU cores occupied by the management resources on each NUMA node is balanced. Therefore, the number of CPU cores available to the virtual machine is also equal. When a NUMA node creates a virtual machine, a larger size virtual machine can be created. For example, the server in Table 2 can create a virtual machine with a maximum specification of 80CPU and 340G memory.

In other embodiments of the present application, when a computing node adopts a centralized resource allocation strategy, the CPU of the computing node may be allocated for management resources in the following manner:

According to the NUMA attributes corresponding to the first management resource, the second management resource and the third management resource, the first management resource, the second management resource and the third management resource are sequentially allocated to the first NUMA node. That is to say, three types of management resources with different NUMA attributes are allocated to the same NUMA node, and more specifically, it can be allocated to the NUMA node where the management resources that use a certain NUMA node are located.

Of course, if the number of CPU cores occupied by the management resources is relatively large, the remaining CPU cores can be allocated to other NUMA nodes when the allocation cannot be completed on a fixed NUAM node. That is, after the third management resource is allocated to the first NUMA node, if the third management resource still needs to occupy H CPU cores, then H CPU cores are allocated from the second NUMA node or the third NUMA node to the third management resource.

The following describes the process of resource allocation using the centralized resource allocation strategy.

As shown in FIG. 5 , a flowchart of a method for resource allocation based on NUMA attributes provided by an embodiment of the present application. Referring to FIG. 5 , the method includes the following steps:

It should be noted that steps S501 and S502 in the embodiment shown in FIG. 5 are the same as steps S301 and S302 in the embodiment shown in FIG. 3 . For details, please refer to the introduction in the embodiment shown in FIG. 3 , which will not be repeated here. Repeat.

S501: Divide the CPU and memory on the server into Q NUMA nodes.

S502: Allocate the resource L to a fixed NUMA node among the Q NUMA nodes.

S503: Allocate the resource M to the same NUMA node as the resource L.

In this embodiment of the present application, when a centralized resource allocation strategy is used, it is necessary to allocate management resources to a NUMA node as much as possible. Therefore, after resource L is allocated, resource M can be allocated to the same NUMA node as resource L node. Exemplarily, assuming that resource L is allocated to NUMA0 first, when resource M is allocated, resource M may also be allocated to NUMNA0.

S504: Allocate the resource N to the same NUMA node as the NUMA node where the resource L and the resource M are located.

For the resource N, the resource N can be preferentially allocated to the NUMA nodes occupied by the resource L and the resource M, so as to ensure that the management resources can be centrally allocated to the same NUMA node.

S505: Determine whether there are remaining unallocated resources N. If there is remaining resource N, continue to step S506.

If the number of CPU cores occupied by resource N is less than the set threshold, that is, when the number of CPU cores is relatively small, resource N may be able to be allocated to the NUMA nodes occupied by resource L and resource M.

If the number of CPU cores occupied by resource N is greater than the set threshold, that is, when the number of CPU cores is relatively large, resource N cannot be allocated completely, then resource N will be left over, that is, there will be unallocated resource N that needs to be continued. distribute.

S506: Allocate the remaining resource N to other NUMA nodes except the fixed NUMA node among the Q NUMA nodes, until the resource N is allocated.

In this embodiment of the present application, if there is a remaining unallocated resource N after the resource N is allocated to the NUMA node where the resource L and the resource M are located, the remaining resource N may be allocated to another NUMA node. If another NUMA node can allocate the remaining resource N, it ends; if the other NUMA node still has not allocated the remaining resource N, continue to allocate the remaining resource N to other NUMA nodes until the resource N is allocated.

Through the above steps, management resources can be centrally allocated to one NUMA node, so that when creating a virtual machine, the number of virtual machines configured on a single NUMA node can be increased.

Further, as a possible implementation manner, in this embodiment of the present application, a configuration policy can also be customized, that is, the user can specify the NUMA nodes occupied by the resource M and the resource N as well as the CPU core occupied by the NUMA node according to requirements. number.

The resource allocation strategy in the embodiment shown in FIG. 5 will be introduced below with reference to specific examples.

Table 1 is still used as the allocation form before the centralized resource allocation strategy is applied. From Table 1, according to the resource allocation in Table 1, the maximum specification of a single NUMA virtual machine that a user can create is: 23CPU, that is, the corresponding NUMA2 node. The virtual machine is the maximum size.

After utilizing the centralized resource allocation strategy in the present application, the allocation of management resources by NUMA nodes on the server can be referred to as shown in Table 3.

table 3

It can be seen from Table 3 that after the centralized allocation strategy is used, the CPU cores occupied by management resources are concentrated in a NUMA node. Therefore, when a single-NUMA virtual machine is created, a larger-sized virtual machine can be created. For example, the server in Table 3 can create a virtual machine with a maximum specification of 24CPU. In other words, the number of CPU cores is more.

Based on the above embodiments, the present application further provides a resource allocation device based on NUMA attributes. Referring to FIG. 6 , the device 600 may include: an acquisition unit 601 and an allocation unit 602 .

Wherein, the obtaining unit 601 is used for obtaining the resource allocation strategy and the NUMA attribute of the management resource issued by the cloud management platform; the allocation unit 602 is used for obtaining the resource allocation strategy and the management resource according to the obtaining unit 601 The NUMA attribute of the central processing unit CPU of the computing node is allocated to the management resource.

In a possible design, when the resource allocation strategy includes a balanced resource allocation strategy, the allocation unit 602 is specifically configured to, according to the resource allocation strategy and the NUMA attribute of the management resource, be configured as follows: The management resource is allocated to the central processing unit CPU of the computing node: according to the NUMA attributes corresponding to the first management resource, the second management resource and the third management resource, respectively, the first management resource, The second management resource and the third management resource are allocated to the first NUMA node, the second NUMA node and the third NUMA node.

In a possible design, the allocating unit 602 is further configured to: after the third management resource is allocated to the third NUMA node, if the third management resource still needs to occupy H CPU cores, then Allocate H CPU cores from the first NUMA node and the second NUMA node with the most remaining CPU cores to the third management resource, or allocate H/2 CPU cores from each of the first NUMA node and the second NUMA node to all the CPU cores. the third management resource, or allocate H CPU cores from the fourth NUMA node to the third management resource.

In a possible design, when the resource allocation strategy includes a centralized resource allocation strategy, the allocation unit 602 is specifically configured to, according to the centralized resource allocation strategy and the NUMA attribute of the management resource, as follows: Allocate the CPU of the computing node to the management resource: according to the NUMA attributes corresponding to the first management resource, the second management resource, and the third management resource, sequentially assign the first management resource, all the The second management resource and the third management resource are allocated to the first NUMA node.

In a possible design, the allocating unit 602 is further configured to: after the third management resource is allocated to the first NUMA node, if the third management resource still needs to occupy H CPU cores, then The H CPU cores are allocated from the second NUMA node or the third NUMA node to the third management resource.

Wherein, all relevant contents of the steps involved in the above method embodiments can be cited in the functional descriptions of the corresponding functional modules, which will not be repeated here.

The division of modules in the embodiments of this application is schematic, and is only a logical function division. In actual implementation, there may be other division methods. In addition, each functional module in each embodiment of this application may be integrated into one processing unit. In the device, it can also exist physically alone, or two or more modules can be integrated into one module. The above-mentioned integrated modules can be implemented in the form of hardware, and can also be implemented in the form of software function modules.

FIG. 7 shows a NUMA attribute-based resource allocation apparatus 700 provided in an embodiment of the present application. The apparatus 700 includes at least one processor 702 for implementing or supporting the apparatus 700 in implementing the diagrams provided in the embodiments of the present application. 6 shows the function of the distribution unit. Exemplarily, the processor 702 may allocate the central processing unit CPU of the computing node, etc. to the management resource according to the resource allocation policy and the NUMA attribute of the management resource obtained by the obtaining unit 601. For details, refer to: The detailed description in the method example will not be repeated here.

The apparatus 700 may also include at least one memory 701 for storing program instructions. Memory 701 and processor 702 are coupled. The coupling in the embodiments of the present application is an indirect coupling or communication connection between devices, units or modules, which may be in electrical, mechanical or other forms, and is used for information exchange between devices, units or modules. The processor 702 may cooperate with the memory 701 . Processor 702 may execute program instructions and/or data stored in memory 701 . At least one of the at least one memory may be included in the processor.

The apparatus 700 may also include a communication interface 703 for communicating with other devices over a transmission medium. The processor 702 may utilize the communication interface 703 to send and receive data.

The present application does not limit the specific connection medium between the communication interface 703 , the processor 702 and the memory 701 . In the embodiment of the present application, in FIG. 7 , the memory 701 , the processor 702 , and the communication interface 703 are connected through a bus 704 , and the bus is represented by a thick line in FIG. 7 . The bus can be divided into an address bus, a data bus, a control bus, and the like. For ease of presentation, only one thick line is used in FIG. 7, but it does not mean that there is only one bus or one type of bus.

In this embodiment of the present application, the processor 702 may be a general-purpose processor, a digital signal processor, an application-specific integrated circuit, a field programmable gate array or other programmable logic device, a discrete gate or transistor logic device, or a discrete hardware component, which can implement Alternatively, each method, step, and logic block diagram disclosed in the embodiments of the present application are executed. A general purpose processor may be a microprocessor or any conventional processor or the like. The steps of the methods disclosed in conjunction with the embodiments of the present application may be directly executed by a hardware processor, or executed by a combination of hardware and software modules in the processor.

In this embodiment of the present application, the memory 701 may be a non-volatile memory, such as a hard disk drive (HDD) or a solid-state drive (SSD), etc., or a volatile memory (volatile memory), For example RAM. Memory is, but is not limited to, any other medium that can be used to carry or store desired program code in the form of instructions or data structures and that can be accessed by a computer. The memory in this embodiment of the present application may also be a circuit or any other device capable of implementing a storage function, for storing program instructions.

Optionally, the computer-executed instructions in the embodiment of the present application may also be referred to as application code, which is not specifically limited in the embodiment of the present application.

Embodiments of the present application further provide a computer-readable storage medium, including instructions, which, when executed on a computer, cause the computer to execute the method of the foregoing embodiment.

Embodiments of the present application also provide a computer program product, including instructions, which, when executed on a computer, cause the computer to execute the methods of the above embodiments.

The embodiment of the present application further provides a chip, and the logic in the chip is used to execute the method of the above embodiment.

The embodiments of the present application are described with reference to flowcharts and/or block diagrams of methods, apparatuses (systems), and computer program products according to the embodiments of the present application. It will be understood that each flow and/or block in the flowcharts and/or block diagrams, and combinations of flows and/or blocks in the flowcharts and/or block diagrams, can be implemented by instructions. These instructions may be provided to the processor of a general purpose computer, special purpose computer, embedded processor or other programmable data processing device to produce a machine such that execution of the instructions by the processor of the computer or other programmable data processing device produces a A means for the functions specified in a flow or flows of a flowchart and/or a block or blocks of a block diagram.

These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory result in an article of manufacture comprising instruction means, the instructions The apparatus implements the functions specified in the flow or flow of the flowcharts and/or the block or blocks of the block diagrams.

These computer program instructions can also be loaded on a computer or other programmable data processing device to cause a series of operational steps to be performed on the computer or other programmable device to produce a computer-implemented process such that The instructions provide steps for implementing the functions specified in the flow or blocks of the flowcharts and/or the block or blocks of the block diagrams.

Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.

Claims

A NUMA attribute-based resource allocation method, wherein the method is applied to a computing node, and the method includes:

Obtain the resource allocation policy issued by the cloud management platform and the NUMA attributes of the management resources;

The central processing unit CPU of the computing node is allocated to the management resource according to the resource allocation policy and the NUMA attribute of the management resource.
The method of claim 1, wherein the management resources include a first management resource, a second management resource, and a third management resource;

The NUMA attributes of the first management resource include:

Need to allocate the CPU of the first NUMA node of the computing node to the first management resource;

The NUMA attributes of the second management resource include:

The CPU of any NUMA node of the computing node can be allocated to the second management resource;

The NUMA attributes of the third management resource include:

CPUs may be allocated to the third management resource in the compute nodes across NUMA nodes.
The method of claim 2, wherein the resource allocation strategy comprises a balanced resource allocation strategy, and the computing node is allocated to the management resource according to the resource allocation strategy and the NUMA attribute of the management resource The central processing unit CPU, including:

Allocate the first management resource, the second management resource, and the third management resource according to the NUMA attributes corresponding to the first management resource, the second management resource, and the third management resource, respectively. to the first NUMA node, the second NUMA node, and the third NUMA node.
The method of claim 3, wherein the method further comprises:

After the third management resource is allocated to the third NUMA node, if the third management resource still needs to occupy H CPU cores, the first NUMA node and the second NUMA node with the most remaining CPU cores are allocated from the one with the most remaining CPU cores. H CPU cores are allocated to the third management resource, or H/2 CPU cores are allocated from each of the first NUMA node and the second NUMA node to the third management resource, or H CPU cores are allocated from the fourth NUMA node to the third management resource.
The method of claim 2, wherein the resource allocation strategy comprises a centralized resource allocation strategy, and the management resource is allocated the management resource according to the centralized resource allocation strategy and a NUMA attribute of the management resource. The CPU of the compute node, including:

According to the NUMA attributes corresponding to the first management resource, the second management resource and the third management resource respectively, the first management resource, the second management resource and the third management resource are sequentially allocated to the first NUMA node.
The method of claim 5, wherein the method further comprises:

After the third management resource is allocated to the first NUMA node, if the third management resource still needs to occupy H CPU cores, then H CPU cores are allocated from the second NUMA node or the third NUMA node to the third NUMA node. the third management resource.
A resource allocation device based on NUMA attributes, comprising:

an obtaining unit, the obtaining unit is used to obtain the resource allocation strategy issued by the cloud management platform and the NUMA attribute of the management resource;

an allocation unit, configured to allocate the central processing unit CPU of the computing node to the management resource according to the resource allocation policy and the NUMA attribute of the management resource obtained by the obtaining unit.
The apparatus of claim 7, wherein the management resources include a first management resource, a second management resource, and a third management resource;

The attributes of the first management resource include:

Need to allocate the CPU of the first NUMA node of the computing node to the first management resource;

The attributes of the second management resource include:

The CPU of any NUMA node of the computing node can be allocated to the second management resource;

The attributes of the third management resource include:

CPUs may be allocated to the third management resource in the compute nodes across NUMA nodes.
The apparatus according to claim 8, wherein when the resource allocation strategy includes a balanced resource allocation strategy, the allocation unit is specifically configured to use the resource allocation strategy and the NUMA of the management resource in the following manner attribute, the central processing unit CPU of the computing node is allocated to the management resource:

Allocate the first management resource, the second management resource, and the third management resource according to the NUMA attributes corresponding to the first management resource, the second management resource, and the third management resource, respectively. to the first NUMA node, the second NUMA node, and the third NUMA node.
The apparatus of claim 9, wherein the distribution unit is further used for:

After the third management resource is allocated to the third NUMA node, if the third management resource still needs to occupy H CPU cores, the first NUMA node and the second NUMA node with the most remaining CPU cores are allocated from the one with the most remaining CPU cores. H CPU cores are allocated to the third management resource, or H/2 CPU cores are allocated from each of the first NUMA node and the second NUMA node to the third management resource, or H CPU cores are allocated from the fourth NUMA node to the third management resource.
The apparatus according to claim 8, wherein when the resource allocation strategy includes a centralized resource allocation strategy, the allocation unit is specifically configured to manage resources according to the centralized resource allocation strategy and the management resource in the following manner NUMA attribute to allocate the CPU of the compute node for the management resource:

According to the NUMA attributes corresponding to the first management resource, the second management resource and the third management resource respectively, the first management resource, the second management resource and the third management resource are sequentially allocated to the first NUMA node.
The apparatus of claim 11, wherein the distribution unit is further configured to:

After the third management resource is allocated to the first NUMA node, if the third management resource still needs to occupy H CPU cores, then H CPU cores are allocated from the second NUMA node or the third NUMA node to the third NUMA node. the third management resource.
A computer-readable storage medium, characterized in that the storage medium stores computer instructions, which, when executed by a computer, cause the computer to execute the method according to any one of claims 1-6 .
A computer program product, characterized in that the computer program product comprises computer instructions, which, when executed by a computer, cause the computer to perform the method according to any one of claims 1-6.