Browse Prior Art Database

Dynamic VM creation for ensuring quality of server (QoS) guarantees in a system-x virtualisation environment

IP.com Disclosure Number: IPCOM000178349D
Original Publication Date: 2009-Jan-23
Included in the Prior Art Database: 2009-Jan-23
Document File: 3 page(s) / 88K

Publishing Venue

IBM

Abstract

This disclosure proposes a method for ensuring Quality of Service (QoS) guarantees in a Virtual Machine (VM) environment.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 52% of the total text.

Page 1 of 3

Dynamic VM creation for ensuring quality of server (QoS) guarantees in a system-x virtualisation environment

Authors -- Pradipta Kumar, Vikas Bhardwaj

According to this method resources to a virtual machine are allocated by taking into account the expected average load. The resources allocated to the virtual machines are based on the prior load analysis of applications resource requirements. Whenever the load on a VM increases beyond a certain threshold (for eg. because of increase in the number of transactions to an application server hosted on this VM) another copy of this VM (with the same application server) is started for load balancing. The load balancer is made aware of the additional VM and it routes the requests to both the VMs so as to balance the load. This new virtual machine can be hosted on any of the physical machine which has free available resources to host the VM.

Many such new VMs can be created to handle increase in load,

provided enough resources are

No assumption is required for the peak load, ever increasing peak limit can also be

satisfied by creating a new VM whenever the load crosses a certain predefined threshold
2. Resources of a physical server are effectively utilized without giving unnecessary buffer resources to any VM. The VMs are optimally sized for a particular work load.

The disclosure is explained in more details with the following embodiments.

A virtual machine (say VM-1) is created with application App-1 installed and 8 GB of RAM allocated to it. VM-1 is able to handle up to 2000 transactions per second which is an average load for the server. At the time of peak load the transactions to the application increases up to 3000 transactions per second so at that time another virtual machine (say VM-2 which is a copy of VM-1) with App-1 installed is created for load balancing. Based on the prior analysis it is known that 4 GB of RAM is required to handle 1000 transactions per second and consequently 4GB RAM is allocated to the VM-2.

Moreover if after some duration peak load reaches up to 4000 transactions per second then third virtual machine VM-3 with 4 GB of RAM can be created to load balance application App-1.

If number of transactions are less then or equal to 2000

per second only then additional virtual

machines VM-2 and VM-3 can be destroyed or kept in power off mode.

Prior analysis to find out resource requirement for a particular work load is already well known.

available on the physical servers to host the VMs.

An alternate approach is to create a new VM with increased resources and routing all new transactions to the new VM. As for the existing VM, once it is done with the processing of the older transactions, it is sh...