Method for rapidly configuring virtual machine resources in cloud computing data center

Abstract

The present invention discloses a method for rapidly configuring virtual machine resources in a cloud computing data center. The method comprises: a resource controller receives a virtual machine job request from a first application server and responds to the request: if the request is received, a physical server cluster which is optimal in queue is selected according to a job type of the request and a virtual machine job is issued to the physical server cluster, and if the request is rejected, reject information is returned; the physical server cluster receivs the virtual machine job from the resource controller and enqueues the job according to tolerable latest job completion time; and when a dispatching cycle of the physical server cluster is reached, a resource optimization strategy of sequentially dispatching in a non-preemptive queue is adopted to dispatch the virtual machine job. One single virtual machine job is sequentially subjected to the steps; and virtual machine jobs in different steps can be processed simultaneously. According to the method for rapidly configuring the virtual machine resources in the cloud computing data center, which is provided by the present invention, the configuration speed of the virtual machine resources is improved, so that the data center can meet service quality requirements of big data cloud computing.

Description

technical field [0001] The invention relates to the field of cloud computing, in particular to a method for rapidly configuring virtual machine resources in a cloud computing data center. Background technique [0002] Cloud computing is a new computing model and service model. The cloud computing data center provides computing resources such as CPU and memory of physical servers and storage resources such as hard disks to users in a leased manner through virtualization technology and cluster technology. Tenants can dynamically request according to their scale of big data processing and storage. resource. In the cloud computing data center, the resource request of each tenant is abstracted into one or more virtual machine (VirtualMachine) jobs. Each virtual machine job corresponds to a certain amount of cloud computing infrastructure resources (including CPU, memory, and hard disk, etc.). A physical server cluster can run multiple virtual machines at the same time. [0003...

AI Technical Summary

Problems solved by technology

[0004] However, in the actual application process, the main problem existing in the virtual machine resource scheduling method of the cloud computing data center in the prior art is that the service quality of the virtual machine jobs cannot be guaranteed: in a sudden period, there are generally thousands to tens of thousands of virtual machine jobs Requests flood into the data center at the same time. If the scheduling management node spends too much time in the process of resource allocation for each job, it will prolong the waiting delay of all jobs waiting to be scheduled, resulting in virtual machine job timeouts and data center resource utilization. low

However, the scheduling management node is required to quickly select a priority scheduling job from the queue and then select one of thousands of physical servers to configure resources for it in a very short scheduling decision period. Processor and memory access operations, as well as the communication of control signaling between scheduling management nodes and physical servers, pose strict requirements. However, the difficulty of current hardware implementation makes it the bottleneck of current high-speed cloud computing data centers.

Although the processing method of the existing technology that immediately rejects the job request because there is no physical server matching the resource specification of the job at the current moment reduces the time consumed by the scheduling management node scheduling decision to a certain extent, this method also reduces the Business income and worsened user experience. In fact, a search / computing resource application generally only needs a short virtual machine job running time, and the job completion time limit (Deadline) of these applications often allows it to tolerate a certain queuing delay

Images