CN108519917A - A resource pool allocation method and device - Google Patents

Abstract

The invention discloses a kind of resource pool distribution method and devices.This method includes：According to the resource information of hardware resource, hardware resource is divided into different types of logical resource pond；According to the type of task, by the logical resource pond of the mission dispatching to corresponding types, and run on the hardware resource in the logical resource pond.The present invention first carries out resource pool division to physical hardware resources, and when dividing resource pool, making each resource pool tool, there are one types, in the task of distribution, task is distributed into corresponding resource pool according to type, the feature for having accomplished perception resource sharing platform and task makes to combine the two over time and space, effectively increases the resource utilization of resource sharing platform.

Description

A resource pool allocation method and device

technical field

The present invention relates to the field of computer technology, in particular to a resource pool allocation method and device.

Background technique

A typical resource sharing platform is a data center (Data Center), such as: Amazon Cloud, Google Cloud, Alibaba Cloud, etc. Each data center has tens of thousands of multi-brand or multi-series servers, and these servers form a resource sharing platform through interconnection.

In the era of big data, there are usually millions of tasks running at the same time on the resource sharing platform. Today, when resources are limited, resource sharing platforms use oversold resources to avoid waste of resources. Take Figure 1 as an example to illustrate the working principle of oversold resources. Resource overselling means that on a single server, the sum of resources allocated to tasks by the resource sharing platform is greater than the capacity of the server. Since the resources occupied by the task are much smaller than the resource allocation in part of the time, even if the sum of the allocation of these tasks is greater than the server capacity, the server can still meet the resource requirements of the task when the task is executed. With resource oversold, the number of tasks executed on the server at the same time is greater than that of non-oversold.

In the existing technology, different tasks have different requirements for resources, and different resources (such as different server brands or series) have a great impact on the same task. Of course, even the same task may have different resource requirements at different times. However, although the resource overselling method can avoid resource waste to a certain extent, it cannot reasonably allocate resources according to the task requirements, and cannot effectively improve resource utilization.

Contents of the invention

The technical problem to be solved by the present invention is to provide a resource pool allocation method and device to solve the problem of low resource utilization in the prior art.

In order to solve the above-mentioned technical problems, the present invention solves through the following technical solutions:

The present invention provides a method for allocating resource pools, including: dividing hardware resources into different types of logical resource pools according to the resource information of the hardware resources; and sending the tasks to corresponding logical resource pools according to the types of tasks , and run on hardware resources in the logical resource pool.

Among them, according to the resource information of the hardware resources, the hardware resources are divided into different types of logical resource pools, including: setting up multiple oversold resource pools with different oversold ratios; according to the performance information of the hardware resources, the hardware resources are divided into different types. oversold resource pool.

Wherein, according to the resource information of the hardware resources, the hardware resources are divided into different types of logical resource pools, including: in each oversold resource pool, according to the configuration information and performance information of the hardware resources, the hardware resources are divided into the super Sell different sub-resource pools in the resource pool.

Wherein, after sending the task to the logical resource pool of the corresponding type, it also includes: monitoring the tasks in the logical resource pool; The task is migrated to hardware resources meeting preset target conditions in other logical resource pools.

Wherein, the types include: overselling type, offline type, online type, interactive type, calculation-intensive type, access-intensive type, and high input/output type.

The present invention also provides a device for allocating resource pools, including: a dividing module, used to divide hardware resources into different types of logical resource pools according to the resource information of the hardware resources; The task is delivered to a logical resource pool of a corresponding type, and runs on hardware resources in the logical resource pool.

Wherein, the division module is configured to: set a plurality of oversold resource pools with different oversold ratios; and divide the hardware resources into different oversold resource pools according to the performance information of the hardware resources.

Wherein, the division module is further configured to: in each oversold resource pool, divide the hardware resources into different sub-resource pools in the oversold resource pool according to the configuration information and performance information of the hardware resources.

Wherein, the device further includes a migration module; the migration module is configured to monitor the tasks in the logical resource pool after the task is sent to the corresponding type of logical resource pool; when the task satisfies the preset When migrating the condition, the task running on the hardware resource is migrated to a hardware resource satisfying the preset target condition in another logical resource pool.

Wherein, the types include: overselling type, offline type, online type, interactive type, calculation-intensive type, access-intensive type, and high input/output type.

The beneficial effects of the present invention are as follows:

The present invention firstly divides physical hardware resources into resource pools, and when dividing resource pools, makes each resource pool have a type, and when assigning tasks, assigns tasks to corresponding resource pools according to types, achieving a perceptual resource sharing platform And the characteristics of the task, combining the two in time and space, effectively improving the resource utilization of the resource sharing platform.

Description of drawings

Fig. 1 is a working principle diagram of resource overselling mode in the prior art;

FIG. 2 is a flowchart of a resource pool allocation method according to a first embodiment of the present invention;

Fig. 3 is a flow chart of dividing steps of hardware resources according to the second embodiment of the present invention;

FIG. 4 is a schematic diagram of division of hardware resources according to a second embodiment of the present invention;

FIG. 5 is a schematic diagram of division of hardware resources according to a second embodiment of the present invention;

Fig. 6 is a structural diagram of a resource pool allocation device according to a third embodiment of the present invention.

Detailed ways

The present invention will be described in further detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described here are only used to explain the present invention, not to limit the present invention.

Embodiment one

This embodiment provides a resource pool allocation method. Fig. 2 is a flowchart of a resource pool allocation method according to the first embodiment of the present invention.

Step S210, divide the hardware resources into different types of logical resource pools according to the resource information of the hardware resources.

Resource information, including but not limited to: CPU brand, CPU speed, CPU utilization, memory capacity, memory utilization, and network bandwidth. Wherein, resource information that changes as the task runs can be used as performance information, and resource information that does not change as the task runs can be used as configuration information.

The types include, but are not limited to: oversold, offline, online, interactive, computation-intensive, access-intensive, and high I/O (Input/Output, input/output) types.

Specifically, the division manner may be set according to specific requirements. Hardware resources with the same resource information can be divided into the same type of logical resource pool. For example: Divide hardware resources with a content capacity of 1T into access-intensive logical resource pools. Hardware resources that meet preset type conditions can be divided into logical resource pools corresponding to the type conditions. For example: set a speed threshold, and hardware resources whose CPU speed is greater than the speed threshold are divided into computing-intensive logical resource pools.

Step S220, according to the type of the task, deliver the task to a logical resource pool of a corresponding type, and run on the hardware resources in the logical resource pool.

The type of task corresponds to the type of logical resource pool.

In this embodiment, after the task is delivered to the logical resource pool of the corresponding type, the tasks in the logical resource pool are monitored; when the task meets the preset migration condition, the hardware resource is run The task is migrated (transferred) to a hardware resource satisfying a preset target condition in another logical resource pool.

Migration conditions can be set according to specific needs. For example, the migration condition is that the performance of the task is less than a preset threshold, and the hardware resources in the logical resource pool where the task is located cannot meet the performance of the task.

The target condition can be set according to specific requirements. For example, the target condition is that the hardware resources in the logical resource pool can make the performance of the task greater than or equal to a preset threshold.

In other words, the migration condition means that the performance of the task is less than the preset threshold, and the logical resource pool where the task is located cannot improve the performance of the task above the preset threshold; the target condition is that the task is migrated to the target logical resource pool The resource can boost the task performance above a preset threshold.

For example: if the number of floating-point operations (performance) of a CPU-intensive task is less than 1G FLOPS/s (preset threshold), and the CPU of the logical resource pool (such as a Pentium CPU) is busy (the average CPU utilization is 90%), then determine The task satisfies the preset migration condition, and the task can be migrated to a high-performance logical resource pool whose CPU utilization of hardware resources is less than 60% (target condition), such as a Xeon server cluster.

In this embodiment, before sending the task to the logical resource pool of the corresponding type, a virtual machine or a container (container) can be set on the hardware resource of the logical resource pool, and after the task is sent to the hardware resource, the Tasks run on virtual machines or containers. When a task needs to be migrated, only the virtual machine or container where the task is running needs to be migrated.

In this embodiment, the logical resource pool includes a plurality of hardware resources, and before sending the task to the logical resource pool of the corresponding type, the hardware that can run the task can be determined in the logical resource pool according to the load balancing algorithm resource, and then deliver the task to the determined hardware resource.

In this embodiment, physical hardware resources are first divided into resource pools, and when dividing resource pools, each resource pool has a type. When assigning tasks, the tasks are assigned to corresponding resource pools according to types, so as to achieve perceptual resource sharing. The characteristics of the platform and tasks combine the two in terms of time and space, effectively improving the resource utilization of the resource sharing platform.

Embodiment two

This embodiment describes the division of hardware resources. In this embodiment, logical resource pools of different levels are set to obtain a more fine-grained logical resource pool.

Fig. 3 is a flowchart of the steps of dividing hardware resources according to the second embodiment of the present invention.

Step S310, setting multiple oversold resource pools with different oversold ratios.

Since the performance of hardware resources is different, the oversold ratio that can be tolerated is different. In order to further improve resource utilization, multiple oversold resource pools are set up, and the oversold ratio is set for each oversold resource pool.

Step S320, divide the hardware resources into different oversold resource pools according to the performance information of the hardware resources.

For example: allocate hardware resources with good performance to oversold resource pools with a large oversold ratio, and divide hardware resources with poor performance to oversold resource pools with a small oversold ratio.

Step S330, in each oversold resource pool, divide the hardware resources into different sub-resource pools in the oversold resource pool according to the configuration information and performance information of the hardware resources.

The child resource pool is a more fine-grained logical resource pool divided in each oversold resource pool (parent resource pool). Each child resource pool has its corresponding type.

Types of sub-resource pools, including but not limited to: offline, online, interactive, computing-intensive, access-intensive, and high I/O.

In this embodiment, one or more parent resource pools and one or more child resource pools may be set as required.

In this embodiment, each logical resource pool (oversold resource pool and sub-resource pool) corresponds to one or more sets of server sets (hardware resources), and logical resource pools may correspond to the same server.

As shown in Figure 4 and Figure 5, in the resource sharing platform, the resource pool of oversold area 1, the resource pool of oversold area 2 and the resource pool of oversold area 3 are set; , set the computing-intensive resource pool, access-intensive resource pool, and high I/O resource pool; servers 1 to 3 are allocated to the computing-intensive resource pool, and servers 1 to 3 are allocated to the access-intensive resource pool, that is, the computing-intensive resource pool The hardware resources and the hardware resources of the access-intensive resource pool overlap, and 4 SSD (SolidState Drives, solid state hardware) servers are divided into the high I/O resource pool.

The hierarchical logical division of resource pools in this embodiment makes resource pool processing tasks more targeted, and assigns different tasks to different resource pools, thereby improving task processing performance and assigning tasks with complementary resource requirements to the same server , providing concurrency of task processing, reducing resource conflicts, and further improving resource utilization.

In this embodiment, when assigning a task, according to the type of the task, a logical resource pool corresponding to the type is found. In this embodiment, because tasks can be executed in the manner of oversold resources, any task can run in any oversold resource pool.

Online tasks have high requirements on response performance, and resources are always occupied;

Offline tasks consume a lot of resources and have no special performance requirements, but they need to be completed before the specified time point;

Interactive tasks require high response performance, but resources are only occupied when users use them;

Computation-intensive tasks require high resource computing performance;

Access-intensive tasks require high resource memory capacity;

High I/O tasks have a high amount of access to interfaces.

Since various tasks have resource occupation time differences, this embodiment, after identifying the task type, sends the task to the logical resource pool of the corresponding type for processing, which can greatly improve the processing speed and resource utilization.

When the performance of the hardware resource running the task is insufficient, the task can be migrated between child resource pools, or directly between parent resource pools, as shown in Figure 4 and Figure 5 .

Embodiment three

This embodiment provides a resource pool allocation device. Fig. 6 is a structural diagram of a resource pool allocation device according to a third embodiment of the present invention.

The resource pool allocation device includes:

The division module 610 is configured to divide the hardware resources into different types of logical resource pools according to the resource information of the hardware resources.

The delivery module 620 is configured to deliver the task to a logical resource pool of a corresponding type according to the type of the task, and run on the hardware resources in the logical resource pool.

Optionally, the division module 610 is configured to: set multiple oversold resource pools with different oversold ratios; and divide the hardware resources into different oversold resource pools according to the performance information of the hardware resources.

Optionally, the division module 610 is further configured to: in each oversold resource pool, divide the hardware resources into different sub-resource pools in the oversold resource pool according to the configuration information and performance information of the hardware resources .

Optionally, the device further includes a migration module (not shown in the figure); the migration module is configured to, after sending the task to a logical resource pool of a corresponding type, monitor the Task: when the task satisfies a preset migration condition, migrate the task running on the hardware resource to a hardware resource satisfying a preset target condition in another logic resource pool.

Optionally, the types include: overselling type, offline type, online type, interactive type, calculation-intensive type, access-intensive type, and high input/output type.

The functions of the device described in this embodiment have been described in the method embodiments shown in Figs. I won't go into details here.

Although preferred embodiments of the present invention have been disclosed for illustrative purposes, those skilled in the art will appreciate that various modifications, additions and substitutions are possible, and therefore, the scope of the present invention should not be limited to the above-described embodiments.

Claims (10)

1. A resource pool allocation method, characterized in that, comprising: According to the resource information of the hardware resources, divide the hardware resources into different types of logical resource pools; According to the type of the task, the task is delivered to a logical resource pool of a corresponding type, and runs on hardware resources in the logical resource pool. 2. The method according to claim 1, wherein, according to the resource information of the hardware resources, the hardware resources are divided into different types of logical resource pools, including: Set up multiple oversold resource pools with different oversold ratios; According to the performance information of the hardware resources, the hardware resources are divided into different oversold resource pools. 3. The method according to claim 2, wherein, according to the resource information of the hardware resources, the hardware resources are divided into different types of logical resource pools, including: In each oversold resource pool, the hardware resources are divided into different sub-resource pools in the oversold resource pool according to the configuration information and performance information of the hardware resources. 4. The method according to claim 1, further comprising: after sending the task to a logical resource pool of a corresponding type: monitoring tasks in the logical resource pool; When the task satisfies a preset migration condition, the task running on the hardware resource is migrated to a hardware resource satisfying a preset target condition in another logic resource pool. 5. The method according to any one of claims 1-4, wherein the types include: oversold, offline, online, interactive, computing-intensive, access-intensive, high input/ output type. 6. A resource pool allocation device, characterized in that it comprises: The division module is used to divide the hardware resources into different types of logical resource pools according to the resource information of the hardware resources; The delivery module is configured to deliver the task to a logical resource pool of a corresponding type according to the type of the task, and run on the hardware resources in the logical resource pool. 7. The device according to claim 6, wherein the dividing module is configured to: Set up multiple oversold resource pools with different oversold ratios; According to the performance information of the hardware resources, the hardware resources are divided into different oversold resource pools. 8. The device according to claim 7, wherein the dividing module is further configured to: In each oversold resource pool, the hardware resources are divided into different sub-resource pools in the oversold resource pool according to the configuration information and performance information of the hardware resources. 9. The device according to claim 6, further comprising a migration module; The migration module is configured to monitor the tasks in the logical resource pool after sending the tasks to the logical resource pool of a corresponding type; when the tasks meet preset migration conditions, upload the hardware resources to The running task is migrated to hardware resources meeting preset target conditions in other logical resource pools. 10. The device according to any one of claims 6-9, wherein the types include: oversold, offline, online, interactive, computing-intensive, access-intensive, high input/ output type.