Jul 09, 2009 (06:07 AM EDT)
VMware Discloses Load Balancing In vSphere 4
Read the Original Article at InformationWeek
VMware said Wednesday that an enhanced feature of its vSphere 4 virtual machine management software can improve application performance by up to 47%.
The Distributed Resource Scheduler in vSphere 4 monitors virtual machine usage across a server cluster to see whether the virtual machine is loaded on the server that best meets its needs. DRS can perform load balancing in the cluster to ensure the best possible application performance.
VMware conducted lab tests with the revised DRS to see how well it coud balance a load of both lightly and heavily utilized copies of Microsoft SQL Server databases on a vSphere 4 managed cluster of four hosts. The latest version of DRS shifted workloads among the servers until they showed a 47% improvement in results, said Stephen Herrod, VMware's CTO.
The intelligence now included in DRS indicates that mission-critical business applications can be put on virtual servers and still perform at a level consistent with their business role, he said.
"DRS optimizes efficiency while providing guaranteed levels of performance. This allows customers to maximize the potential of their data center resources," he said in the announcement.
Application performance in virtual machines remains something of a black art for many early implementers.
Instead of overprovisioning an application with extra, raw server CPU and memory, system administrators are being called on to save resources by virtualizing the application on a shared server, while guaranteeing its continued successful business operation. Traditional systems management systems can confirm whether the application is running but can't tell what user response times are.
VMwares DRS was first launched in 2006 as part of vSphere's predecessor product, Virtual Infrastructure 3. VSphere 4 was announced last September at VMworld in Las Vegas and became available April 22. It included many feature updates of Virtual Infrastructure 3 and added new ones aimed at managing the data center as a fully virtualized resource.
The announcement included a statement from Fazil Habibulla, VP and system engineer at Natizis Capital Markets, that his firm had been able to nearly double its server consolidation to eight to 10 per ESX Server host with the improved application performance. The increased number of virtual machines were run on the same hardware as the four to five VMs per physical server that preceded them.
The same announcement quoted another customer, Brian Doyle, network specialist for Jenner & Block, as saying DRS's automated monitoring and load balancing saved administrative time formerly spent watching for virtual machine overload.