Browse Prior Art Database

Intelligent memory resource allocation for virtualization environment based on historical monitoring data

IP.com Disclosure Number: IPCOM000200447D
Publication Date: 2010-Oct-14
Document File: 8 page(s) / 85K

Publishing Venue

The IP.com Prior Art Database

Abstract

This invention describes one resource optimizer in virtualization environment which uses the historical monitoring data, the business priorities to optimize the resource allocation dynamically and get the maximum performance within the limited available physical resources.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 49% of the total text.

Page 01 of 8

Intelligent memory resource allocation for virtualization environment based on historical monitoringdata

With Virtualization infrastructures , such as vmware, virtual machines share the physical memory on the host server . Virtualization infrastructure allows memory over commitment (the sum of memory setting of each vm is greater than the real physical memory size of the host), which is good for resource reuse and sharing . However, with memory over commitment technology , the vm running on the host might not get real physical memory , when it claims for memory . In this situation, the virtual memory, which is simulated from hard disk will be used instead, and this introduces the potential performance degradation , because of the corresponding page swapping . For the high priority VM, virtualization infrastructure provides ways to reserved an preset amount of physical memory , which is restrictively used by this VM, even when the actual used memory is lesser than the reserved memory . It is very important to set appropriate reserved memory size of the VMs running on the same host server to get the resource use optimization and to reduce resource waste .

However, existing technologies only provide static settings of the reserved memory , and this presents two challenges .
1. During sometime (for example, many requests come in a short interval ), the static setting is toosmall , and therefore the high priority VM (and the application running on this VM ) has to compete with other VMs running on the same host for memory , and might not get the physical memory.
2. During sometime (for example, the request number is small ), the static setting is too large , andtherefore the reserved memory is not use efficiently and other VMs can not obtain it (because it's reserved restrictively ). It is a waste.

The priority of the VM is changed by times as well . For example, during the working hours , for banking applications , the counter application has higher priority than online banking application , however, during the office closing time , it's the opposite.

The root cause is that the reserved settings should not be static . They should be changed based on real use of the memory size and the business priorities.

We can easily get the following data :
1. The virtualization environment is monitored and we can get the history of total memory usage (including physical and virtual memory ) of each virtual machine by time .
2. The business priority of each VM by time (for example by hour), which is set by users.
3. We can also get the total available memory in the physical host .

This invention describes one resource optimizer in virtualization environment which uses the historical monitoring data , the business priorities to optimize the resource allocation dynamically and get the maximum performance within the limited available physical resources.

The memory usage of the each VM can be recorded as the following diagram , which is the summarizat...