Method and system for self-adaptive on-demand resource allocation in a virtualized environment

Abstract

The invention discloses a system for adaptively allocating resources as required in a virtualization environment. The system comprises a dynamic perception request distribution module, a 1-physical machine (PM):N-virtual machine (VM) module and a data center global management module, wherein the 1-PM:N-VM module allocates the resources on a PM according to user experiences which are collected in real time; the dynamic perception request distribution module distributes loads to proper VMs according to monitored application request load information and VM volume information and responds to requests; and the data center global management module judges whether the VMs are required to migrate between the PMs to be re-placed according to collected PM resource load information, and judges whether a new PM is released or applied to an idle resource pool to quit or enter application service when the PM is excess or insufficient. The invention also discloses a method for adaptively allocating the resources as required in the virtualization environment. The method comprises an adaptive VM dynamic volume perception request distribution strategy, a 1-PM:N-VM resource allocation strategy and a VM migration strategy. The invention has an application prospect in the technical field of computers.

Description

(1) Technical field [0001] The invention relates to resource allocation, in particular to a method and system for self-adaptive on-demand resource allocation in a virtualized environment, and belongs to the technical field of computers. (2) Background technology [0002] With the rapid development of Internet applications, more and more service providers outsource their services to data centers, making the scale of data centers larger and more complex. Virtualization technology is playing an increasingly important role in building large-scale data centers due to its following characteristics: 1. Improve resource utilization of infrastructure. A physical machine (PM) can run multiple virtual machines (VM) at the same time, and the virtual machine monitor (VMM) provides more fine-grained resource allocation mechanism support, such as the default credit of xen (an open source virtualization product). Scheduling policy provides mechanism support for dynamic allocation of CPU re...

AI Technical Summary

Problems solved by technology

Adjustment based on resource usage has the following disadvantages: (a) With the development of Internet applications, applications become more and more "big" and more complex, and the definition of expected resource usage becomes more and more appropriate. difficulty

If the definition is too high, it will affect the performance of the application

If the definition is too low, it will lead to waste of resources

(b) Accurate sampling of resource usage becomes more difficult in a virtualized environment due to the overhead of filtering the hypervisor

(c) When multiple virtual machines are running on the same physical machine, the accuracy of predicting the real resource demand based on the current measured resource usage rate is very poor

[0007] 3) The traditional request distribution strategy for the physical machine environment is no longer applicable in the virtualized environment

However, in reality, web services are very bursty, and when this happens, adding new physical machines to participate in the running time will cost too much

According to statistics, if the system fails to respond to user requests on time and exceeds user satisfaction, the service provider will not only face the risk of "increasing" the user's frustration with the service, what's more, the user may turn to competitors' services, which will affect the company. products and company reputation, causing great economic losses

Images