WO2018072687A1

WO2018072687A1 - Resource scheduling method and apparatus, and filtered scheduler

Info

Publication number: WO2018072687A1
Application number: PCT/CN2017/106553
Authority: WO
Inventors: 朱旭琪
Original assignee: 华为技术有限公司
Priority date: 2016-10-19
Filing date: 2017-10-17
Publication date: 2018-04-26
Also published as: CN107968802B; CN107968802A; CN111258737A; CN111258737B

Abstract

The present invention relates to a resource scheduling method and apparatus, and a filtered scheduler. The method comprises: a first filtered scheduler acquiring a scheduling request message carrying service requirements, converting the scheduling request message into a scheduling task message, and notifying each filtered scheduler in a scheduling system by means of asynchronous communication, wherein the scheduling task message is used for instructing the filtered scheduler to determine, according to the service requirements, a scheduling result in a virtual group associated with the filtered scheduler, and to send the scheduling result to the first filtered scheduler; acquiring a scheduling result determined, according to the scheduling task message, by the filtered scheduler in the scheduling system; and when a pre-set condition is satisfied, determining, according to a pre-set algorithm, a final scheduling result in the acquired scheduling result, the final scheduling result comprising one or more servers satisfying the service requirements. By means of the method, the processing capacity of scheduling and the scheduling efficiency of the whole scheduling system are improved.

Description

Method, device and filter scheduler for resource scheduling

Technical field

The present invention relates to the field of communications, and in particular, to a resource scheduling method, apparatus, and filter scheduler in the field of virtualization management.

Background technique

With the widespread application of virtualization technology, how to use the cloud computing management platform to better realize the unified management and scheduling of resources has received more and more attention. One of the important technologies of the management platform is resource scheduling, which is the business. Requests are allocated to appropriate resources, including computing, storage, networking, etc., to improve resource utilization.

In the cloud computing management platform, the most commonly used scheduler is a filtering scheduler, that is, when receiving a scheduling request, first filtering out all the server sets satisfying the business requirements according to the business requirements carried in the scheduling request, and then All servers in the server set are sorted according to a predefined weight calculation method, and the scheduler selects the server with the highest weight to process the service demand. However, with the increasing business scenarios in large data centers, the single filter scheduler is responsible for the scheduling of all resources in the virtualized system, which cannot meet the business requirements, resulting in serious performance bottlenecks. More and more data centers adopt Multiple filter schedulers implement scheduling requests in parallel to share the scheduling tasks of the entire large-scale data center.

In the prior art, a plurality of filter schedulers are used in the scheduling system to form a scheduler cluster. When there are multiple scheduling requests to be processed in the message queue, the filter scheduler in the scheduler cluster separately obtains different scheduling requests, and The scheduling result is determined according to different scheduling requests, but for the same scheduling request, a filtered scheduler first filters out the server set that satisfies the scheduling request in all servers in the scheduling system, and then according to the predefined rights. The value calculation method calculates the weights of all the servers in the server set, and then sorts the servers according to the weights, and finally selects one or more servers as the scheduling result of the scheduling request. In the large-scale data center, the scheduling method is The same filter scheduler is required to filter and sort all servers, resulting in long processing time and low efficiency of the same scheduled task. On the other hand, the available resources of each server in the prior art are recorded and saved by the database, but the server only informs the database to update the available resources of the server when the business requirements in the scheduling request are completed. That is to say, the available resource information of the server recorded in the database is non-real-time information, and the server determined by the filter scheduler according to the information may not meet the service requirement, thereby affecting the judgment of the filter scheduler and increasing the scheduling failure. Probability. Therefore, the method for parallel scheduling of multiple filter schedulers in the prior art has the problems that the scheduling request processing time is long, the efficiency is low, and the misjudgment causes the service processing to fail, which affects the resource scheduling capability and resource allocation efficiency of the scheduling system.

Summary of the invention

The embodiment of the invention provides a method, a device and a filter scheduler for resource scheduling, which can solve the problem that the scheduling system processes the scheduling request for a long time, the efficiency is low, and the misjudgment causes the service processing to fail, and the scheduler system is improved. Resource scheduling capabilities and resource allocation efficiency.

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

In a first aspect, a method for resource scheduling is provided, where the method is applied to a scheduling system, where the scheduling system includes at least two filter schedulers and at least two servers, wherein each filter scheduler is associated with one virtual group a group, each virtual group includes all servers that can be scheduled by the filter scheduler associated with the virtual group, and one server belongs to only one virtual group; first, the first filter scheduler obtains a scheduling request message, The scheduling request message carries a service requirement, and the first filtering scheduler is any filtering scheduler in the scheduling system; then, the first filtering scheduler converts the scheduling request message into a scheduling task message, and asynchronously The communication mode notifies each of the filter schedulers in the scheduling system, wherein the scheduled task message carries the service requirement and the role information of the first filter scheduler, where the schedule task message is used to indicate that the filter scheduler is in accordance with the service requirement. Determining the scheduling result in the virtual group associated with the filter scheduler, and transmitting the scheduling result to the first filter And the scheduling result includes one or more servers that meet the service requirement; next, the first filtering scheduler obtains a scheduling result determined by the filtering scheduler in the scheduling system according to the scheduling task message; when the preset condition is met The first filter scheduler determines a final scheduling result in the obtained scheduling result according to a preset algorithm, where the final scheduling result includes one or more servers that meet the service requirement.

It is worth noting that each virtual group includes one or more servers.

Optionally, the scheduling task message may further include identifier information and IP address information of the first filter scheduler.

Optionally, in the scheduling task message processing process, the first filtering scheduler that obtains the scheduling request message may record the role information of each filtering scheduler in the current scheduling task message by using the data table in the specified file or the database. For example, the role of the filter scheduler may include both Leader and Worker, wherein the filter scheduler of the Worker role is responsible for filtering the associated virtual in the set of at least one server that it can schedule. The group determines one or more servers that meet the service requirements as the scheduling result; the filter scheduler of the Leader role is a filter scheduler that obtains the scheduling request message, and is responsible for converting the scheduling request message into a scheduling task message, and notifying the scheduling system. All the filter schedulers have scheduling task messages to be processed, summarize the scheduling results determined by the filter scheduler of all the worker roles, and select one or more services satisfying the service requirements among all the summarized scheduling results according to the preset algorithm. For the final scheduling results.

It is worth noting that the role information is temporary information, that is, it is only valid during the processing of the current scheduling request message. When multiple scheduling request messages exist in the system, after different filtering schedulers obtain scheduling task messages, each scheduling request message is determined by a leader role of the filtering scheduler to determine the final scheduling result, and each type of role Only one message can be processed at a time, that is, the filter role of the Leader role can only process one scheduling request message at a time, and the filter scheduler of the Worker role can only process one scheduling task message at a time. After the scheduling request message processing ends, the filtering scheduler releases the role information corresponding to the current scheduling request message processing.

In this embodiment, each of the filter schedulers in the scheduler system determines the scheduling result in the respective set of servers that can be scheduled according to the same scheduling request, and sends the scheduling result to the filtering schedule for acquiring the scheduling request message. When the preset condition is met, the filter scheduler determines one or more servers as the final scheduling result among all the summarized scheduling results, and has only one filtering schedule for the same scheduling request message in the prior art. Compared with the scheduling result, the parallel processing of multiple filtering schedulers solves the problem of long processing time and low efficiency of scheduling tasks in the prior art, and improves the resource scheduling capability and resource allocation efficiency of the scheduling system.

In a possible implementation manner, the notifying, by the asynchronous communication manner, each filtering scheduler in the scheduling system includes: the first filtering scheduler storing the scheduling task message in a message queue, so that Every A filter scheduler listens to the message queue and acquires the scheduled task message.

It is worth noting that after the scheduling request message in the message queue is removed by any filter scheduler, the message queue software deletes the scheduling request message in the message queue to avoid waste of resources caused by repeated processing. The scheduler cannot obtain the scheduling request message again. The scheduling task persists during the processing of the scheduling request until the first filter processes the scheduling task, and then notifies the message queue software to delete the scheduling task message, so that all the filtering schedulers in the storage cluster can be asynchronous. The way to get the scheduled task. After obtaining the scheduling task, the filtering scheduler in the storage cluster can determine the server in the virtual group corresponding to each filtering scheduler that is most suitable for performing the business requirements in the scheduling task.

Optionally, the scheduling task message may also be stored in the database, and the manner in which the first filtering scheduler obtains the scheduling request message may also be performed by listening to the database, that is, when there is a scheduling request message to be processed in the database, the slave database Obtain the scheduling request message.

In another possible implementation manner, the notifying, by the asynchronous communication manner, each filtering scheduler in the scheduling system includes: the first filtering scheduler storing the scheduling task message in a database, so that Each of the filter schedulers listens to the database and acquires the scheduled task message.

In another possible implementation manner, the filtering scheduler determines, according to the service requirement, a scheduling result in a virtual group associated with the filtered scheduler, including: the filtering scheduler obtaining the Scheduling a task message; the filter scheduler determines, in the virtual group associated with the filter scheduler, a server that satisfies the service requirement according to the service requirement; and the filter scheduler performs a pair of predefined weight algorithm The determined servers are sorted, and one or more are sequentially selected in order as the scheduling result.

In another possible implementation manner, the preset conditions include:

Obtaining, by the first filter scheduler, the number of the scheduling results meets a first threshold; or

The first filter scheduler obtains that the number of the scheduling results meets a first threshold and the receiving time satisfies a second threshold.

It should be understood by those skilled in the art that each filter scheduler can only process one scheduled task message at a time. When multiple scheduling task messages need to be processed simultaneously in the scheduling system, for the same filter scheduler, if current If the scheduled task request message is not processed for a long time, the filtered scheduler cannot acquire and process a new scheduled task message. To avoid the influence of the previous scheduling task processing process on the current scheduling task, the first filtering scheduler is When the final scheduling result is determined, the first threshold may be set. When the number of the summarized scheduling results meets the first threshold, one or more servers are selected as the final scheduling result in the obtained scheduling result.

In another possible implementation manner, the method for determining, by the first filtering scheduler, the final scheduling result in the scheduling result according to a preset algorithm includes any one of the following manners:

The first filter scheduler sorts all servers in the scheduling result according to the weight determined by the predefined weight algorithm, and sequentially selects one or more servers as the final scheduling result; or

The first filter scheduler sequentially selects one or more servers from the scheduling result in the order of receiving time as the final scheduling result; or

The first filter scheduler randomly selects one or more servers from the scheduling result as the final scheduling result;

The predefined weight algorithm may be based on at least a CPU, a remaining amount of memory, a remaining amount of the disk, a network traffic, a CPU type, a disk type, a network card type, and a number of times the disk reads and writes per second. One defines different weights and calculates the weight of each server.

It is worth noting that after determining the final scheduling result, the first filtering scheduler notifies the message queue software to delete the processed scheduling task message. At the same time, the first filter scheduler also sends the optimal scheduling result to the task assignment node, and the task assignment node sends the service requirement in the scheduling task to the server corresponding to the final scheduling result, and the server performs the business requirement. The specific implementation process.

In summary, for a scheduling request message, each filtering scheduler in the scheduling system divides and cooperates, and respectively determines scheduling results in respective associated virtual groups, and a filtering scheduler summarizes the filtering scheduler according to the scheduling task message. Determining the scheduling result, and determining one or more servers as the final scheduling result in the foregoing scheduling result, compared with the prior art, the same scheduling request message is processed by multiple filtering schedulers in parallel, reducing scheduling Processing time improves the processing efficiency of the scheduling system. On the other hand, each filter scheduler uses resource preemption information to record the resource information required for the determined scheduling result, thereby solving the scheduling failure problem caused by the fact that the record server resource occupancy cannot be updated in real time in the database in the prior art, thereby Improved the ability of the scheduler cluster to manage new resources in large data.

In a second aspect, an apparatus for resource scheduling is provided, the apparatus comprising various modules for performing the resource scheduling method of the first aspect or any of the possible implementations of the first aspect.

In a third aspect, a method for resource scheduling is provided, where the method is applied to a scheduling system, where the scheduling system includes at least two filter schedulers and at least two servers, wherein each filter scheduler is associated with one virtual group a group, each virtual group includes all servers that can be scheduled by the filter scheduler associated with the virtual group, and one server belongs to only one virtual group; first, the first filter scheduler obtains a scheduling request message, The scheduling request message carries a service requirement, and the first filtering scheduler is any filtering scheduler in the scheduling system; then, the first filtering scheduler converts the scheduling request message into a scheduling task message, and the scheduling is performed. The task message is sent to the server in the scheduling system except the first filter scheduler, where the scheduling task message carries the service requirement and the role information of the first filter scheduler, and the schedule task message is used to indicate the scheduling system. Other filter schedulers other than the first filter scheduler are associated with other filter schedulers based on business requirements. Determining a scheduling result in the quasi-group, and transmitting the scheduling result to the first filtering scheduler; next, the first filtering scheduler acquires a scheduling result determined by the filtering scheduler in the scheduling system according to the scheduling task message; When the preset condition is met, the first filter scheduler determines the final scheduling result in the acquired scheduling result according to the preset algorithm, and the final scheduling result includes one or more servers that meet the service requirement.

Through the description of the foregoing content, the first filtering type and other filtering schedulers in the scheduling system directly communicate. For the same scheduling request message, each filtering scheduler in the scheduling system divides and cooperates, and the filtering of the scheduling request message is obtained. The scheduler determines the final scheduling result, reduces the processing time, and improves the processing efficiency of the scheduling request message.

In a possible implementation manner, the preset conditions include:

In another possible implementation manner, the method for determining, by the first filtering scheduler, the final scheduling result in the acquired scheduling result according to a preset algorithm includes any one of the following manners:

The weight of the first filter scheduler determined according to the predefined weight algorithm is used in the scheduling result There are servers to sort, and one or more servers are selected in turn for the final scheduling result; or

The predefined weight algorithm may be based on at least one of a CPU remaining amount of the server, a remaining memory amount, a remaining amount of the disk, a network traffic, a CPU type, a disk type, a network card type, and a number of times the disk reads and writes per second. Define different weights and calculate the weight of the obtained server.

Through the description of the foregoing content, for a scheduling request message, each filtering scheduler in the scheduling system divides and cooperates, determines scheduling results in respective associated virtual groups, and summarizes all scheduling results by a filtering scheduler and determines final scheduling. As a result, compared with the prior art, for the same scheduling request message, multiple filtering schedulers are processed in parallel, which reduces the scheduling processing time and improves the processing efficiency of the scheduling system. On the other hand, each filter scheduler uses resource preemption information to record the resource information required for the determined scheduling result, thereby solving the scheduling failure problem caused by the fact that the record server resource occupancy cannot be updated in real time in the database in the prior art, thereby Improved the ability of the scheduler cluster to manage new resources in large data.

In a fourth aspect, a filter scheduler is provided, where the filter scheduler includes a processor, a memory, a communication interface, and a bus, and the processor, the memory, and the communication interface are connected by a bus and complete communication with each other. The memory is configured to store computer execution instructions, and when the filter scheduler is running, the processor executes a computer execution instruction in the memory to perform the first aspect and the first with hardware resources in the filter scheduler Any of the aspects described in the possible implementations.

A fifth aspect, a computer readable medium for storing a computer program, the computer program comprising instructions for performing the method of the first aspect and the possible implementation of any of the first aspects.

In summary, for a scheduling request, each of the filter schedulers in the scheduler cluster cooperates and determines the scheduling result in each corresponding virtual group, and is obtained by the filter scheduler that obtains the scheduling request. The scheduling result determines that one or more servers are final scheduling results, which avoids resource competition problems caused by multiple filtering schedulers working in parallel in the prior art. On the other hand, each filter scheduler uses the resource camping information to record the resource information required for the determined scheduling result, and solves the problem of scheduling failure caused by the inability of the record server resource occupancy in the database in the prior art to be updated in real time. The dispatcher cluster resource allocation message improves the ability of the scheduler cluster to manage new resources in large data.

DRAWINGS

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings to be used in the embodiments of the present invention will be briefly described below. It is obvious that the drawings described below are only some embodiments of the present invention, Those skilled in the art can also obtain other drawings based on these drawings without paying any creative work.

FIG. 1 is a schematic diagram of a scheduling system according to an embodiment of the present invention;

2 is a schematic flowchart of a method for resource scheduling according to an embodiment of the present invention;

FIG. 3 is a schematic flowchart diagram of another method for resource scheduling according to an embodiment of the present disclosure;

FIG. 4 is a schematic flowchart diagram of another method for resource scheduling according to an embodiment of the present disclosure;

FIG. 5 is a schematic flowchart diagram of another method for resource scheduling according to an embodiment of the present disclosure;

FIG. 6 is a schematic diagram of an apparatus for resource scheduling according to an embodiment of the present invention;

FIG. 7 is a schematic diagram of a filter scheduler according to an embodiment of the present invention.

detailed description

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

1 is a schematic diagram of a scheduling system provided by the present invention. As shown in the figure, the scheduling system includes a management node, a task allocation node, a filter scheduler, a server, and a database, wherein two or more filters are used. The scheduler forms a scheduler cluster.

The management node may pre-create a virtual group for each filter scheduler in the scheduling system, and the virtual group includes all servers that the filter scheduler can schedule, that is, the virtual group includes one or more servers, and A server belongs to only one virtual group.

For example, as shown in FIG. 1 , there are three filter schedulers in the scheduling system. The management node creates one virtual group for each filter scheduler in advance, including:

The virtual group 1 is a virtual group associated with the filter scheduler 1, and the virtual group includes the server 1 and the server 2.

The virtual group 2 is a virtual group associated with the filter scheduler 2, which includes the server 3 and the server 4.

The virtual group 3 is a virtual group associated with the filter scheduler 3, which includes the server 5 and the server 6.

The correspondence between the filter scheduler, the virtual group, and the server can utilize database records as shown in FIG. The database is also used to record the used resource information and available resource information of each server, for example, the usage and availability of the server's CPU, memory, disk, network card and other resources. When the filtered scheduler processes the scheduling request, the filtered scheduler can query its associated virtual group and the available resource information of each server in the virtual group through the database.

Optionally, the filter scheduler may also record the correspondence between the filter scheduler, the virtual group, and the server by using a specified file.

As a possible embodiment, when creating a virtual group, the management node may divide the server into different virtual groups according to attributes such as an identification (ID), an operating system type, and a hard disk type.

It is worth noting that the number of servers in each virtual group can be the same or different. In the following detailed description of the present invention, three types of filter schedulers are included in the scheduler cluster, and two virtual servers in the virtual group corresponding to each filter scheduler are exemplified in detail.

Optionally, when there is a faulty server in the virtual group or a new server exists in the scheduling system, the management node may adjust the number of servers in each virtual group according to a predefined adjustment policy. For example, if there is a faulty server in the virtual group 1, the management node can adjust each virtual according to the priority order of the filter scheduler. The number of servers in the quasi-group; if a new filter scheduler is added to the scheduling system, the management node can add the server to a virtual group associated with the filter scheduler according to the processing capability of each filter scheduler. in.

Alternatively, the virtual group may also be managed by the task assignment node, that is, the task assignment node creates a virtual group for each filter scheduler and dynamically adjusts the number of servers in the virtual group.

It should be noted that the management node and the task allocation node in FIG. 1 may be deployed in the same server, or may be deployed in different servers separately, which is not limited by the present invention. In the following description of the present invention, the management node and the filter scheduler shown in FIG. 1 are deployed in the same server as an example for detailed description.

Alternatively, the virtual group may also be managed by a third party management device or device outside the scheduling system that is capable of learning all of the filtered scheduler and server information.

In the scheduling system, the task allocation node is further configured to receive the scheduling request message, and store the scheduling request message carrying the service requirement in a message queue (Message Queue, MQ), where each filtering scheduler in the scheduling system is monitored. In the mode, if there is a scheduling request message to be processed in the message queue, if the scheduling request message is to be processed, the scheduling request message is obtained, and the final scheduling result is determined, and then the final scheduling result is sent to the message. The task assignment node is sent by the task assignment node to the one or more servers corresponding to the final scheduling result.

The message queue is a container for saving messages during the transmission of the message. In the scheduling system shown in FIG. 1, the server 100 is used to deploy the message queue software, and the message queue software can use the memory of the server 100 to save and delete the task allocation node and the filtering method. Messages passed between the schedulers, and messages transmitted between all filter schedulers, wherein the server 100 includes at least one message queue, and each message queue can store different types of messages according to an initial configuration, for example, in a storage system. There are two message queues. Message queue 1 is used to save messages transmitted between the task assignment node and the scheduling cluster, such as scheduling request messages and final scheduling results. Message queue 2 is used to save messages transmitted between the filter schedulers, such as The task message is scheduled and the scheduling result determined by each filter scheduler. Optionally, the message queue software may also save the message in the form of a message queue cluster. Each message queue cluster includes at least two message queues, and different message queue clusters are used to store different types of messages. For example, there are two in the storage system. Message queue cluster, message queue cluster 1 is used to save messages transmitted between the task assignment node and the scheduler cluster. For example, one message queue in message queue cluster 1 is used to save the scheduling request message, and another message queue is used to save the final scheduling result. Message queue cluster 2 is used to save messages transmitted between the filter schedulers. For example, one message queue in message queue cluster 2 is used to save scheduling task messages, and another message queue is used to save the schedule determined by each filter scheduler. result.

It is worth noting that during the implementation process, the message queue can be deployed by a separate server, or multiple servers can be deployed simultaneously, and server resources can be shared with other types of software. In the following description of the present invention, a scenario in which a message queue is deployed by a separate server as shown in FIG. 1 is taken as an example for further elaboration.

Optionally, the communication mode between the task allocation node and the filtering scheduler in the scheduling system and all the filtering schedulers may also be implemented by using other asynchronous communication methods such as a database, such as a Red is database.

Next, a resource scheduling method provided by an embodiment of the present invention is further introduced in conjunction with FIG. 2, as shown in the figure, the method includes:

S201. The first filter scheduler acquires a scheduling request message.

The scheduling request message carries the service requirement, and the first filtering scheduler is any filtering type in the scheduling system. scheduler. The manner in which the first filter scheduler obtains the scheduling request message may be a manner of listening to the message queue, that is, when there is a scheduling request message to be processed in the message queue, the scheduling request message is obtained from the message team.

Optionally, the task assignment node may also send the scheduling request message directly to the first filtered scheduler.

S202. The first filtering scheduler converts the scheduling request message into a scheduling task message, and notifies each filtering scheduler in the scheduling system by asynchronous communication.

Specifically, the scheduling task message carries the service requirement and the role information of the first filtering scheduler, and the task scheduling message is used to indicate that each filtering scheduler in the scheduling system determines the scheduling result in the associated virtual group according to the service requirement. The scheduling result is summarized into a first filtered scheduler, and the scheduling result includes one or more servers that meet the business requirements.

The first filtering scheduler notifies each filtering scheduler that there is a scheduling task message to be processed in the scheduling system, and the scheduling task message may be stored in the message queue, and each filtering scheduler in the scheduling system is listening to the message. When there is a scheduled task message to be processed in the queue, the scheduling task message is obtained.

Optionally, the first filtering scheduler notifies each filtering scheduler that there is a scheduling task message to be processed in the scheduling system, and may also store the scheduling task message in a database used for asynchronous communication, such as a Red is database. All filter schedulers obtain the scheduled task message by periodically listening to the database.

Optionally, the first filtering scheduler notifies each filtering scheduler that there is a scheduling task message to be processed in the scheduling system, and may also obtain an IP address of all filtering schedulers in the scheduling system from the database, according to the IP address. The scheduled task message is sent to all filter schedulers except the first filter scheduler.

S203. The first filter scheduler acquires a scheduling result determined by the filter scheduler in the scheduling system according to the scheduling task message.

Specifically, after determining the scheduling result, the filtering scheduler in the scheduling system determines, according to the role information in the scheduling task message, a message queue used by the first filtering scheduler to receive the scheduling result, and stores the scheduling result in the message queue. The first filter scheduler listens to the message queue, and obtains the scheduling result when there is a new scheduling result in the message queue.

S204. When the preset condition is met, the first filter scheduler determines a final scheduling result in the acquired scheduling result according to a preset algorithm.

Specifically, satisfying the preset condition includes that the quantity of the first filter scheduler acquiring the scheduling result satisfies the first threshold; or the number of the first filter scheduler acquiring the scheduling result satisfies the first threshold and the receiving time meets the second threshold. The final scheduling result determined by the first filter scheduler includes one or more servers that meet the business requirements.

In this embodiment, each of the filter schedulers in the scheduler system determines the scheduling result in the respective set of servers that can be scheduled according to the same scheduling request, and sends the scheduling result to the filtering schedule for acquiring the scheduling request message. When the preset condition is met, the filter scheduler determines one or more servers as the final scheduling result among all the acquired scheduling results, and only one filter scheduler determines the scheduling result in the prior art. In contrast, the parallel processing of multiple filter schedulers solves the problem of long processing time and low efficiency of scheduling tasks in the prior art, and improves resource scheduling capability and resource allocation efficiency of the scheduling system.

The following is a description of the resource scheduling method provided by the embodiment of the present invention. The method includes:

S301. The first filtered scheduler obtains a scheduling request message in a message queue.

After receiving the scheduling request message sent by the application layer, the task allocation node stores the scheduling request message in the message queue, and the first filtering scheduler obtains the scheduling request message, where the first filtering scheduler is a scheduling Any filter scheduler that obtains the scheduling request message in the system. The service requirement carried in the scheduling request message, the service requirement includes information about creating a virtual machine, virtual machine migration information, and virtual machine resource scheduling.

Table 1 shows an example of a message format for a scheduling request message:

Table 1

Specifically, each filtering scheduler in the scheduling system identifies, by means of the interception manner, whether there is a scheduling request message to be processed in the message queue. When there is a scheduling request message to be processed in the message queue, the idle scheduling filter scheduler will The scheduling request is obtained from the message queue. At this time, the first filtering scheduler that obtains the scheduling request message is responsible for processing the scheduling request.

It is worth noting that after the scheduling request message in the message queue is removed by any filter scheduler, the message queue software deletes the scheduling request message in the message queue to avoid waste of resources caused by repeated processing. The scheduler cannot obtain the scheduling request message again.

S302. The first filter scheduler converts the scheduling request message into a scheduling task message, and stores the scheduling task message in a message queue.

Specifically, the first filtered scheduler adds the role information of the first filtered scheduler in the scheduling request message. The scheduling task message is used to indicate that the filtering scheduler in the scheduling cluster determines the scheduling result, and sends the determined scheduling result to the first filtering scheduler. The scheduling task message also carries the service requirement, that is, the service requirement carried in the scheduling request message in step S301.

Table 2 is an example of a message format in which the first filter scheduler converts the scheduling request message in Table 1 into a scheduling task message:

Table 2

It can be seen from Table 1 and Table 2 above that the scheduling request message is converted into a scheduling task message in the message grid. The identifier and role information of the first filter scheduler and the IP address of the first filter scheduler are added. The message format included in the scheduling request message and the scheduling request task message in Table 1 and Table 2 is only an example. In a specific implementation process, a message format of other content may also be included.

For example, in the scheduling system shown in FIG. 1, if the filtering scheduler 1 acquires the scheduling request message, the filtering scheduler 1 records the role information in the processing of the current scheduling request message, and the filtering scheduler 1 is this. The leader of the scheduling request message, the filter scheduler 2 and the filter scheduler 3 are all Worker roles. Alternatively, the filter scheduler 1 may also have a Worker role, then the filter scheduler 1, the filter scheduler 2, and The filter scheduler 3 is responsible for determining the scheduling result in its associated virtual group, and transmitting the determined scheduling result to the filter scheduler 1, and the filter scheduler 1 is responsible for obtaining the filter scheduler according to the scheduled task message. Determine the scheduling results and determine one or more servers as the final scheduling results.

Further, the first filtering scheduler stores the scheduling task message in the message queue, so that the filtering scheduler in the scheduling system can listen to the message queue, obtain the scheduling task message from the message queue, and process the message.

Specifically, the scheduling task message may be saved in the message queue for storing the scheduling request message in step S301, and all the filtering schedulers in the scheduling system identify the scheduling request message and the scheduling task message during the process of listening to the message queue. The identifier distinguishes between different scheduling request messages and scheduling task messages; optionally, the scheduling request message may also be saved by using a separate message queue, and all filtering schedulers in the scheduling system identify whether there are scheduled scheduling task messages by listening to the specified message queue. .

It is to be noted that, in a manner that the scheduling request message is deleted after being acquired by the first filtering scheduler in step S301, the scheduling task is stored in the message queue in step S302, and the scheduling task is in the processing process of the current scheduling request. It persists until the first filter processes the scheduled task, and then notifies the message queue software to delete the scheduled task message, so that all the filtered schedulers in the storage cluster can acquire the scheduled task in an asynchronous manner. After the filtering scheduler in the storage cluster obtains the scheduling task, it can determine that each filtering scheduler is correct. The server in the virtual group that is best suited to perform the business needs in the scheduled task.

Optionally, after the scheduling request message is converted into the scheduling task message, the first filtering scheduler may also save the scheduling task message in a database for asynchronous communication, and each filtering scheduler periodically listens to the database manner. Identify and obtain the scheduled task message to be processed.

Optionally, after converting the scheduling request into the scheduling task message, the first filtering scheduler may also obtain the IP addresses of all the filtering schedulers in the scheduling system from the database, and directly report to each filtering according to the IP address. The dispatcher sends the scheduled task message.

S303. The first filter scheduler obtains the scheduling task message in a message queue, and saves the determined scheduling result in a message queue.

When there is a scheduled scheduling task message in the message queue, the filtering scheduler of each idle state in the scheduling system acquires the scheduling request message and processes it by means of interception. When the first filter scheduler detects that there is a scheduled task message to be processed, and the first filter scheduler is in an idle state, the first filter scheduler obtains the schedule task message and is associated with the virtual group. Select one or more servers in the group that meet the business requirements as the scheduling result.

It is worth noting that each filter scheduler can only process one scheduled task message at a time, that is, each filter scheduler will acquire and process a new schedule in the message queue only after the current scheduled task message is processed. Task message.

Specifically, the method for determining the scheduling result is the same for each filtering scheduler in the scheduling system. FIG. 4 is a flowchart of a method for determining a scheduling result by the filtering scheduler. As shown in the figure, the specific method includes:

S303a. The filter scheduler determines, according to the service requirement, a server that meets the service requirement in the virtual group associated with the filter scheduler.

The server that meets the service requirement is determined by the filter scheduler according to the service requirement and the remaining available resources of the server in the virtual group associated with the filter scheduler, that is, the virtual group associated with the filter scheduler. Identify one or more servers that meet the business needs.

S303b: The filter scheduler calculates the sorted servers according to the predefined weight algorithm calculation, and sequentially selects one or more servers as the scheduling result in sequence.

Specifically, each filter scheduler calculates a weight of a server that satisfies a business requirement according to a predefined weight algorithm, and sorts the server for the weight, and sequentially selects one or more of a sequence from high to low or low to high. The servers serve as the determined scheduling results. The predefined weight algorithm may define at least one of a server resource, a memory resource, a disk remaining amount, a disk read/write times per second, and a network bandwidth according to a customized requirement of a specific service application. The weights are calculated and the weight of the server is calculated.

Optionally, the scheduling result further includes weight information of the server.

For example, if the server that can be scheduled by the filter scheduler 1 includes the server 1 and the server 2, the service requirement carried in the scheduling request message is to create a virtual machine with 2 CPUs, 8 GB memory, 32 G disks, and 1 network card. After screening, it is determined that both server 1 and server 2 meet the specification requirements for creating a virtual machine, and the server that satisfies the business requirement includes two servers; if the business application has customization requirements for the remaining amount of the disk and memory of the server, the predefined right In the value algorithm, the available memory weight of the server is 0.6, and the available disk weight is 0.4. Specifically, if the available memory resource of the server 1 is 10G, the available disk is 100G, the available memory resource of the server 2 is 20G, and the available disk is 100G, and the weight of the server 1 is 46 (0.6×10+0.4×100=46). The weight of the server 2 is 50 (0.5×20+0.4×100=50), and the weights of the servers satisfying the service requirement are the server 2 and the server 1 in descending order, and the filter scheduler 1 determines that the server 1 is The scheduling result of the scheduling request message in the server set that it can schedule. Optionally, the scheduling result further includes the weight information 50 of the server 2.

Table 3 is an example of a message format of a scheduling result in the above embodiment:

table 3

Through the description of steps S303a to S303b, each filter scheduler in the scheduling system selects at least one server that satisfies the service requirement as the scheduling result in the manner in which the server can be scheduled according to the first screening and reordering.

Further, after the scheduling result is determined, the filtering scheduler sends the determined scheduling result to the specified filtering scheduler according to the role information in the scheduling task, for example, to the first filtering scheduler in step S301. The manner in which the filtering scheduler sends the scheduling result to the first filtering schedule may be any one of the following modes:

Manner 1: The filter scheduler stores the determined scheduling result in the message queue according to the role information in the scheduling task message.

The message queue for storing the scheduling result determined by the filter scheduler may be the same as the message queue for saving the scheduling request message in step S301. At this time, the first filter scheduler distinguishes the scheduling request message and the scheduling task by specifying the message format. The message and the scheduling result, wherein the specified message format for determining the scheduling result may be the "scheduling result" in Table 3. The scheduling result may also be directly sent to the message queue of the first filtering scheduler for receiving the scheduling result according to the role information in the scheduling task message.

Manner 2: Send the determined scheduling result to the first filtering scheduler according to the IP address of the first filtering scheduler in the scheduling task message.

It should be understood that, for the first filter scheduler, if there is a Leader role and a Worker role in the current task, when the first filter scheduler is used as the Leader role of the scheduled task message, the first filter type The scheduler may be in a non-idle state, and the scheduled task message may not be the first scheduled task message to be processed in the message queue. To simplify the implementation process, the first filter will save the converted schedule task message directly. In the message queue, when the first filter scheduler is in an idle state, the scheduled task message is re-acquired for processing, and the scheduling result is stored in a message queue for saving the scheduling result or stored in the local storage. Optionally, when the first filter scheduler has a Leader role and a Worker role for scheduling task messages, the scheduled task message may also be stored in the local storage, when the first filter scheduler is idle. In the state, the scheduling task message identifier distinguishes different scheduling task messages, and the locally saved scheduling task message is preferentially processed.

S304. The other filter scheduler in the scheduling system acquires a scheduling task message in the message queue, and stores the determined scheduling result in the message queue.

Specifically, the other filter scheduler is a filter scheduler of any idle state except the first filter scheduler in the scheduler cluster, and other filter schedules exist when there are scheduled task tasks to be processed in the message queue. The device also obtains the scheduling task message, and selects at least one server that meets the service requirement in the associated virtual group as the scheduling result according to the service requirement carried in the scheduling task message. The process of determining the scheduling result by the other filtering scheduler is the same as that described in step S303, and details are not described herein again.

S305. The first filter scheduler obtains a scheduling result determined by each filter scheduler in the scheduler system.

Specifically, the manner in which the first filtering scheduler obtains the scheduling result is the manner in which the filtering scheduler sends the scheduling result in step S303, and the first filtering scheduler may obtain the scheduling result from the message queue, or may receive the filtering scheduling. The scheduling result sent by the device.

S306. When the preset condition is met, the first filter scheduler determines, in the acquired scheduling result, a final scheduling result according to the preset algorithm, where one or more servers satisfying the service requirement are included in the final scheduling result.

The preset condition may be any one of the following methods:

Manner 1: The first filter scheduler summarizes the number of scheduling results to meet the first threshold.

Manner 2: The first filter scheduler receives the number of scheduling results that meet the first threshold and the receiving time meets the second threshold.

Optionally, when determining the final scheduling result, the first filtering scheduler may limit the receiving time in addition to the number of the scheduled scheduling results, when both meet the requirements, that is, in the summarized scheduling result. Choose a final scheduling result.

Optionally, in a specific implementation process, the first filter scheduler may further select a final scheduling result after receiving the scheduling result sent by all the filtering schedulers in the scheduling system, that is, selecting one of all servers in the scheduling system or A plurality of servers satisfying the business requirements are used as the final scheduling result, and the final scheduling result is one or more servers that satisfy the optimal business requirements.

Further, the processing manner that the first filtering scheduler determines the final scheduling result in the scheduling result according to the preset algorithm includes any one of the following manners:

Manner 1: The first filter scheduler sorts all the servers in the obtained scheduling result according to the weight determined by the predefined weight algorithm, and sequentially selects one or more servers that meet the service requirements as the final scheduling result.

For example, Table 4 is an example of the scheduling result summarized by the first filter scheduler. It is assumed that the service requirement is to create one virtual machine, and the first filter scheduler determines the weight according to the predefined weight algorithm to all servers. Enter The row sorting from high to low is server 3, server 5, and server 1. The final scheduling result is server 3, that is, server 3 is the best server that satisfies the business requirements.

Table 4

调度结果Scheduling result	服务器信息server information	权值信息Weight information
调度结果1Scheduling result 1	服务器1Server 1	4646
调度结果2Scheduling result 2	服务器3Server 3	6060
调度结果3Scheduling result 3	服务器5Server 5	5555

The predefined weight algorithm may be the same as the method for determining the weight in step S303b, or the weight of the server resource may be redefined according to the specific application requirement, and the server weight is calculated. The server resource includes a CPU type, a disk type, a network card type, and Available CPU resources, available memory resources, disk remaining amount, network bandwidth, disk read/write times per second, CPU remaining amount, memory remaining amount, disk remaining amount, etc., the final scheduling result is the optimal one to meet the business requirements. Or multiple servers.

Optionally, the predefined weight algorithm may also determine the weight of the server resource according to the service application not paying attention to the resource information, and calculate the weight information of each server. At this time, the first filter scheduler may sequentially select the sequence from low. One or more servers that meet high business needs are the final scheduling results.

Manner 2: The first filter scheduler selects one or more servers as the final scheduling result in the received scheduling result determined by each filter scheduler in sequence.

Manner 3: The first filter scheduler randomly selects one or more servers from the scheduling result as a final scheduling result.

In addition to mode 1, in order to improve the processing efficiency and reduce the processing time of the scheduling system, the second scheduling method may be determined according to the second filtering mode, that is, the first filtering scheduler determines the final scheduling result according to the receiving time sequence or a random manner, and the final scheduling result is obtained. It includes at least one server that meets business needs.

Through the description of step S301 to step S306, when there is a scheduling request message to be processed in the scheduling system, the scheduling request message is obtained by a filtering scheduler, the scheduling request message is converted into a scheduling task message, and the scheduling is performed. The request message is stored in the message queue, and each filter scheduler in the scheduler system processes the scheduled task message, and summarizes the determined scheduling result to the filter scheduler, and the filter scheduler follows the preset algorithm. The optimal one or more servers that meet the business requirements are determined in the aggregated scheduling result as the final scheduling result. Compared with the prior art, for the same scheduling request message, each filtering scheduler in the scheduling system divides and cooperates and performs parallel processing, which solves the problem that the scheduling processing time is long and the efficiency is low in the prior art, and the whole scheduling system is improved. Scheduling capability and resource scheduling efficiency.

In another possible embodiment of the present invention, each filter scheduler may also record the records recorded in the prior art by recording resource camp-on information after determining the scheduling result in the corresponding virtual group. The available resources of the server cannot be updated in real time, which causes the filter scheduler to misjudge the problem of scheduling failure. The specific method is as shown in FIG. 5, and the method includes:

S501. The filter scheduler records resource camping information after determining the scheduling result according to the scheduling task message.

Specifically, the resource camping information is used to record the scheduling result determined by the filter scheduler. For example, if the service requirement carried in the scheduling task message 1 is to create a virtual machine, the specifications of the virtual machine include 2 CPUs, 10 Gbytes of memory, 2 20 G disks, and 1 network card, assuming that the filter scheduler determines the virtual group associated with it. If the server 1 in the group is used as the scheduling result, the filtering scheduler records the above scheduling task message and the scheduling result as resource camping information.

Table 5 shows an example of resource preemption information:

Optionally, the filter scheduler may record resource preemption information by using a specified file, or may use a database to record resource camp-on information.

It is to be noted that the resource pre-emption information recorded by the first filter scheduler is an information identifier recorded by the filter scheduler, and the physical resources are not actually divided in the server corresponding to the first scheduling result.

Optionally, when the resource preemption information is recorded and the scheduling result is not finally determined, if the new scheduling task message is executed again, the filtering scheduler processes the filtering method for the server involved in the resource camping information during the processing. The scheduler may first obtain the available resource information of the server in the database, subtract the resource preemption information recorded by the filter scheduler, determine the remaining resource information of the server, and determine the server according to the remaining resources of the server. Whether it is possible to execute the business requirements in the new scheduled task message.

S502. When the first filter scheduler determines the final scheduling result, the first filter scheduler notifies the filter scheduler of the final scheduling result, so that the filter scheduler clears the recorded resource camp-on information.

After the first filter scheduler determines the final scheduling result, the final scheduling result is notified to all the filter schedulers in the scheduling system, so that the filtered scheduler in the scheduling system can clear the recorded resource camp-on information according to the final scheduling result. The notification mode may store the final scheduling result in the message queue, or obtain the IP addresses of all the schedulers except the first filter scheduler in the database, and send the final scheduling result according to the IP address information.

Further, the process of clearing resource camp-on information includes the following two situations:

Case 1: If the server in the final scheduling result does not belong to the virtual group associated with the filter scheduler, the filter scheduler directly clears the resource camp-on information recorded by the filter scheduler.

Case 2: If the server in the final scheduling result belongs to the virtual group associated with the filter scheduler, the filter scheduler clears the recorded resource camp-on information after the server in the final scheduling result completes the service processing. .

It is worth noting that after the server corresponding to the final scheduling result completes the service requirement, the server notifies the database in the scheduling system to update its resource occupancy, and notifies the server that the corresponding filter scheduler service requirement is successfully processed. The scheduler accordingly clears the recorded resource preemption information.

In summary, for a scheduling request message, each filtering scheduler in the scheduling system divides and cooperates, determines scheduling results in respective associated virtual groups, and summarizes all scheduling results by a filtering scheduler. The final scheduling result is determined. Compared with the prior art, the same scheduling request message is processed by multiple filtering schedulers in parallel, which reduces the scheduling processing time and improves the processing efficiency of the scheduling system. On the other hand, each filter scheduler uses resource preemption information to record the resource information required for the determined scheduling result, thereby solving the scheduling failure problem caused by the fact that the record server resource occupancy cannot be updated in real time in the database in the prior art, thereby Improved the ability of the scheduler cluster to manage new resources in large data.

It should be noted that, for the above method embodiments, for the sake of simple description, they are all expressed as a series of action combinations, but those skilled in the art should understand that the present invention is not limited by the described action sequence, and secondly, Those skilled in the art should also appreciate that the embodiments described in the specification are all preferred embodiments and the actions involved are not necessarily required by the present invention.

Other reasonable combinations of steps that can be conceived by those skilled in the art in light of the above description are also within the scope of the present invention. Secondly, those skilled in the art should also be familiar with the embodiments described in the specification, which are all preferred embodiments, and the actions involved are not necessarily required by the present invention.

A method and a scheduling system for resource scheduling according to an embodiment of the present invention are described in detail above with reference to FIG. 1 to FIG. 5 . A filter scheduler according to an embodiment of the present invention will be described below with reference to FIG. 6 to FIG. 7 . .

Figure 6 is a device 600 for resource scheduling according to an embodiment of the present invention. The device 600 includes an obtaining unit 601, a processing unit 602, and a determining unit 603.

The acquiring unit is configured to acquire a scheduling request message, where the scheduling request message carries a service requirement;

The processing unit is configured to convert the scheduling request message into a scheduling task message, and notify each filtering scheduler in the scheduling system by using an asynchronous communication manner, where the scheduling task message carries the service a demand and a role information of the device, the scheduling task message is used to indicate that the filter scheduler determines a scheduling result in a virtual group associated with the filter scheduler according to the service requirement, and the scheduling result is Sent to the device, the scheduling result includes one or more servers that meet the service requirements;

The obtaining unit is further configured to acquire a scheduling result that is determined by the filtering scheduler in the scheduling system according to the scheduling task message;

The determining unit is configured to determine, in the preset scheduling result, a final scheduling result according to a preset algorithm, where the final scheduling result includes one or more servers that meet the service requirement. .

It should be understood that the apparatus 600 of the embodiment of the present invention may be implemented by an Application Specific Integrated Circuit (ASIC) or a Programmable Logic Device (PLD), and the PLD may be a complex program logic device ( Complex Programmable Logic Device (CPLD), Field-Programmable Gate Array (FPGA), Generic Array Logic (GAL), or any combination thereof. When the processing method of resource scheduling shown in FIG. 2 is implemented by software, the device 600 and each module thereof may also be software modules.

Through the description of the apparatus 600, each of the filter schedulers in the scheduler system determines the scheduling result in the respective set of servers that can be scheduled according to the same scheduling request, and sends the scheduling result to the filtering method of acquiring the scheduling request message. The scheduler selects a final scheduling result among all the scheduled scheduling results when the preset condition is met, and the plurality of filtering is compared with the prior art that only one filtering scheduler determines the scheduling result. Parallel processing of the scheduler solves the problem of long processing time and low efficiency of scheduling tasks in the prior art. The problem is to improve the resource scheduling capability and resource allocation efficiency of the scheduling system.

Optionally, the processing unit notifying, by asynchronous communication, each of the filtering schedulers in the scheduling system includes: storing the scheduling task message in the message queue, so that each of the filtering schedulers Listening to the message queue and obtaining the scheduling task message.

Specifically, in the scheduling system, each filtering scheduler is notified that there is a scheduling task message to be processed, and the scheduling task message may be stored in a message queue, where each filtering scheduler exists in the listening message queue. The scheduled task message is obtained when the task message is scheduled to be processed.

Optionally, the acquiring unit is further configured to obtain the scheduling task message;

The processing unit is further configured to: determine, according to the service requirement, a server that meets the service requirement in a virtual group associated with the filter scheduler; sort the determined server according to a predefined weight algorithm, and One or more servers are sequentially selected in order as the scheduling result.

Optionally, the preset condition includes: the first filter scheduler acquires the number of the scheduling results to meet a first threshold; or

Optionally, the determining unit determines, according to the preset algorithm, the final scheduling result in the acquired scheduling result, including any one of the following manners:

Sorting all servers in the scheduling result according to the weight determined by the predefined weight algorithm, and sequentially selecting one or more servers as the final scheduling result; or

Selecting one or more servers from the scheduling results in sequence according to the receiving time order as the final scheduling result; or

Randomly selecting one or more servers from the scheduling results as the final scheduling result;

The predefined weight algorithm may be based on at least one of a CPU remaining amount of the server, a remaining memory amount, a remaining amount of the disk, a network traffic, a CPU type, a disk type, a network card type, and a number of times the disk reads and writes per second. Define different weights and calculate the weight of each server.

Optionally, the device further includes a recording unit, and the recording unit is configured to: after determining the scheduling result, Recording resource pre-occupation information, where the resource pre-occupation information is used to identify resource information that needs to be occupied by the filtering result determined by the filtering scheduler in a corresponding server;

The processing unit is further configured to notify the filtered scheduler of the final scheduling result after the first filtered scheduler determines the final scheduling result.

In summary, for a scheduling request message, each filtering scheduler in the scheduling system divides and cooperates, determines scheduling results in respective associated virtual groups, and summarizes all scheduling results by a filtering scheduler and determines final scheduling results. Compared with the prior art, for the same scheduling request message, multiple filtering schedulers are processed in parallel, which reduces the scheduling processing time and improves the processing efficiency of the scheduling system. On the other hand, each filter scheduler uses resource preemption information to record the resource information required for the determined scheduling result, thereby solving the scheduling failure problem caused by the fact that the record server resource occupancy cannot be updated in real time in the database in the prior art, thereby Improved the ability of the scheduler cluster to manage new resources in large data.

Apparatus 600 in accordance with an embodiment of the present invention may correspond to performing the methods described in the embodiments of the present invention, and the above and other operations and/or functions of the various units in apparatus 600 are respectively implemented to implement the various methods of FIGS. 1 through 5. The corresponding process, for the sake of brevity, will not be described here.

FIG. 7 is a filter scheduler 700 provided by the present invention. As shown in FIG. 7, the filter scheduler 700 includes a processor 701, a memory 702, a communication interface 703, and a bus 704. The processor 701, The memory 702 and the communication interface 703 are connected and communicate with each other via a bus 704 for storing computer execution instructions. When the filter scheduler 700 is in operation, the processor 701 executes the memory 702. The computer executes instructions to utilize the hardware resources in the filtered scheduler 700 to perform the following operations:

Obtaining a scheduling request message, where the scheduling request message carries a service requirement;

Transmitting the scheduling request message into a scheduling task message, and notifying each of the filtering schedulers in the scheduling system by using an asynchronous communication manner, where the scheduling task message carries the service requirement and the filtered scheduling The role information of the device 700, the scheduling task message is used to instruct the filter scheduler to determine a scheduling result in a virtual group associated with the filter scheduler according to the service requirement, and send the scheduling result to the The filter scheduler 700, the scheduling result includes one or more servers that meet the service requirements;

Obtaining a scheduling result determined by the filtering scheduler in the scheduling system according to the scheduling task message;

When the preset condition is met, the final scheduling result is determined in the acquired scheduling result according to a preset algorithm, where the final scheduling result includes one or more servers that meet the service requirement.

It should be understood that, in the embodiment of the present invention, the processor 701 may be a CPU, and the processor 701 may also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), and off-the-shelf programmable Gate array (FPGA) or other programmable logic device, discrete gate or transistor logic device, discrete hardware components, etc. The general purpose processor may be a microprocessor or the processor or any conventional processor or the like.

The memory 702 can include read only memory and random access memory and provides instructions and data to the processor 701. A portion of the memory 702 can also include a non-volatile random access memory. For example, the memory 702 can also store information of the device type.

The system bus 704 can include a power bus, a control bus, a status signal bus, and the like in addition to the data bus. However, for clarity of description, various buses are labeled as system bus 704 in the figure.

Through the above description, for each scheduling request, the division of each filter scheduler in the scheduler cluster Collaboration, respectively determining the scheduling result in the respective corresponding virtual group, and collecting the results by the first filtering scheduler that obtains the scheduling request, and finally determining the optimal scheduling result, thereby avoiding multiple filtering methods in the prior art. The parallel operation of the scheduler leads to resource competition issues. On the other hand, each filter scheduler uses the resource camping information to record the resource information required for the determined scheduling result, and solves the problem of scheduling failure caused by the inability of the record server resource occupancy in the database in the prior art to be updated in real time. The dispatcher cluster resource allocation message improves the ability of the scheduler cluster to manage new resources in large data.

It should be understood that the filter scheduler 700 according to an embodiment of the present invention may correspond to the resource scheduling apparatus 600 in the embodiment of the present invention, and may correspond to performing corresponding ones in the method shown in FIG. 2 according to an embodiment of the present invention. The above-mentioned and other operations and/or functions of the respective modules in the filter scheduler 700 are respectively implemented in order to implement the respective processes of the respective methods in FIG. 1 to FIG. 5, and are not described herein again for brevity.

In summary, for a scheduling request, each of the filter schedulers in the scheduler cluster cooperates and determines the scheduling result in each corresponding virtual group, and is summarized by the first filtered scheduler that obtains the scheduling request. All the results and finally determine the optimal scheduling result, avoiding the problem of resource competition caused by the parallel operation of multiple filter schedulers in the prior art. On the other hand, each filter scheduler uses the resource camping information to record the resource information required for the determined scheduling result, and solves the problem of scheduling failure caused by the inability of the record server resource occupancy in the database in the prior art to be updated in real time. The dispatcher cluster resource allocation message improves the ability of the scheduler cluster to manage new resources in large data.

Those of ordinary skill in the art will appreciate that the elements and algorithm steps of the various examples described in connection with the embodiments disclosed herein can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the solution. A person skilled in the art can use different methods for implementing the described functions for each particular application, but such implementation should not be considered to be beyond the scope of the present invention.

A person skilled in the art can clearly understand that for the convenience and brevity of the description, the specific working process of the system, the device and the unit described above can refer to the corresponding process in the foregoing method embodiment, and details are not described herein again.

In the several embodiments provided by the present application, it should be understood that the disclosed systems, devices, and methods may be implemented in other manners. For example, the device embodiments described above are merely illustrative. For example, the division of the unit is only a logical function division. In actual implementation, there may be another division manner, for example, multiple units or components may be combined or Can be integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection through some interface, device or unit, and may be in an electrical, mechanical or other form.

The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, may be located in one place, or may be distributed to multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of the embodiment.

In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

The functions may be stored in a computer readable storage medium if implemented in the form of a software functional unit and sold or used as a standalone product. Based on such understanding, the part of the technical solution of the present invention that contributes in essence or to the prior art or the part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method of various embodiments of the present invention. . The foregoing storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, and the like. .

The above is only a specific embodiment of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art can easily think of changes or substitutions within the technical scope of the present invention. It should be covered by the scope of the present invention. Therefore, the scope of the invention should be determined by the scope of the claims.

Claims

A method for resource scheduling, characterized in that the method is applied to a scheduling system, where the scheduling system includes at least two filter schedulers and at least two servers, wherein each filter scheduler is associated with one virtual a group, each virtual group includes all servers that can be scheduled by a filter scheduler associated with the virtual group, and one server belongs to only one virtual group; the method includes:

The first filter scheduler obtains a scheduling request message, where the scheduling request message carries a service requirement, and the first filter scheduler is any filter scheduler in the scheduling system;

The first filter scheduler converts the scheduling request message into a scheduling task message, and notifies each of the filtering schedulers in the scheduling system by using an asynchronous communication manner, where the scheduling task message carries the a service requirement and role information of the first filter scheduler, the schedule task message is used to instruct the filter scheduler to determine a scheduling result in a virtual group associated with the filter scheduler according to the service requirement, And sending the scheduling result to the first filtered scheduler, where the scheduling result includes one or more servers that meet the service requirement;

Obtaining, by the first filter scheduler, a scheduling result determined by the filtering scheduler in the scheduling system according to the scheduling task message;

When the preset condition is met, the first filter scheduler determines a final scheduling result in the acquired scheduling result according to a preset algorithm, where the final scheduling result includes one or more that meet the service requirement. server.
The method according to claim 1, wherein the notifying each of the filtering schedulers in the scheduling system by asynchronous communication comprises:

The first filtered scheduler stores the scheduled task message in a message queue such that each of the filtered schedulers listens to the message queue and acquires the scheduled task message.
The method according to any one of claims 1 to 2, wherein the filter scheduler determines a scheduling result in a virtual group associated with the filter scheduler according to the service requirement, including:

The filtering scheduler obtains the scheduling task message;

The filter scheduler determines, according to the service requirement, a server that meets the service requirement in a virtual group associated with the filter scheduler;

The filter scheduler sorts the determined servers according to a predefined weight algorithm, and sequentially selects one or more servers as the scheduling result in order.
The method according to any one of claims 1 to 3, wherein the satisfying the preset condition comprises:

Obtaining, by the first filter scheduler, the number of the scheduling results meets a first threshold; or

The first filter scheduler obtains that the number of the scheduling results meets a first threshold and the receiving time satisfies a second threshold.
The method according to any one of claims 1 to 4, wherein the method for determining, by the first filter scheduler, the final scheduling result in the acquired scheduling result according to a preset algorithm comprises any of the following manners. One:

The first filter scheduler sorts all servers in the scheduling result according to the weight determined by the predefined weight algorithm, and sequentially selects one or more servers as the final scheduling result; or

The first filter scheduler sequentially selects one or more servers from the scheduling result in the order of receiving time as the final scheduling result; or

The first filter scheduler randomly selects one or more servers from the scheduling result as the final scheduling result;

The predefined weight algorithm may be based on at least one of a CPU remaining amount of the server, a remaining memory amount, a remaining amount of the disk, a network traffic, a CPU type, a disk type, a network card type, and a number of times the disk reads and writes per second. Define different weights and calculate the weight of each server.
The method according to any one of claims 1 to 5, wherein the method further comprises:

After the filtering scheduler determines the scheduling result, the resource pre-occupation information is used to identify that the scheduling result determined by the filtering scheduler needs to be occupied in the corresponding server. Resource information

After the first filtered scheduler determines the final scheduling result, the method further includes:

The first filter scheduler notifies the filter scheduler of the final scheduling result, so that the filter scheduler clears the resource camp-on information.
An apparatus for resource scheduling, wherein the apparatus includes an acquiring unit, a processing unit, and a determining unit;

The acquiring unit is configured to acquire a scheduling request message, where the scheduling request message carries a service requirement;

The processing unit is configured to convert the scheduling request message into a scheduling task message, and notify each filtering scheduler in the scheduling system by using an asynchronous communication manner, where the scheduling task message carries the service a demand and a role information of the device, the scheduling task message is used to indicate that the filter scheduler determines a scheduling result in a virtual group associated with the filter scheduler according to the service requirement, and the scheduling result is Sent to the device, the scheduling result includes one or more servers that meet the service requirements;

The obtaining unit is further configured to acquire a scheduling result that is determined by the filtering scheduler in the scheduling system according to the scheduling task message;

The determining unit is configured to determine, in the preset scheduling result, a final scheduling result according to a preset algorithm, where the final scheduling result includes one or more servers that meet the service requirement. .
The apparatus according to claim 7, wherein the processing unit notifies each of the scheduling schedulers in the scheduling system by asynchronous communication means that:

The scheduling task message is stored in the message queue such that each of the filtering schedulers listens to the message queue and acquires the scheduling task message.
Apparatus according to any one of claims 7 to 8 wherein:

The obtaining unit is further configured to obtain the scheduling task message;

The processing unit is further configured to: determine, according to the service requirement, a server that meets the service requirement in a virtual group associated with the filter scheduler; sort the determined server according to a predefined weight algorithm, and One or more servers are sequentially selected in order as the scheduling result.
The apparatus according to any one of claims 7 to 9, wherein said satisfying preset conditions comprises:

Obtaining, by the first filter scheduler, the number of the scheduling results meets a first threshold; or

The first filter scheduler obtains that the number of the scheduling results meets a first threshold and the receiving time satisfies a second threshold.
The apparatus according to any one of claims 7 to 10, wherein the determining unit determines, according to a preset algorithm, the final scheduling result in the acquired scheduling result, including any one of the following manners:

Sorting all servers in the scheduling result according to the weight determined by the predefined weight algorithm, and sequentially selecting one or more servers as the final scheduling result; or

Selecting one or more servers from the scheduling results in sequence according to the receiving time order as the final scheduling result; or

Randomly selecting one or more servers from the scheduling results as the final scheduling result;

The predefined weight algorithm may be based on at least one of a CPU remaining amount of the server, a remaining memory amount, a remaining amount of the disk, a network traffic, a CPU type, a disk type, a network card type, and a number of times the disk reads and writes per second. Define different weights and calculate the weight of each server.
Apparatus according to any one of claims 7 to 11, wherein said apparatus further comprises a recording unit;

The recording unit is configured to: after the determining the scheduling result, record resource pre-occupation information, where the resource pre-occupation information is used to identify that the scheduling result determined by the filtering scheduler needs to be occupied in a corresponding server. Resource information;

The processing unit is further configured to notify the filtered scheduler of the final scheduling result after the first filtered scheduler determines the final scheduling result.
A filter scheduler, wherein the filter scheduler includes a processor, a memory, a communication interface, and a bus, and the processor, the memory, and the communication interface are connected by a bus and complete communication with each other. In the memory for storing computer execution instructions, when the filter scheduler is running, the processor executes computer execution instructions in the memory to perform claims 1 to 6 using hardware resources in the filter scheduler Any of the methods described.