Quoting from ComputerWorld
Virtualization lets companies more efficiently use their data center hardware. But just as important is better utilization of the human resources that manage those data centers.
To lighten the load, VMware Inc. in Palo Alto, Calif., released its Distributed Resource Scheduler last October as part of its VirtualCenter 3 management software. DRS automatically balances the workload among pools of virtual servers, allowing servers to operate at 80% or greater utilization.
Once policies are established, DRS manages performance, automatically provisions virtual servers, balances loads and activates fail-over without administrator action. Developers spent about three years testing and refining DRS.
"The more that companies use virtualization in production, the more that DRS will be used, or at least looked at, to manage resource allocation of virtual machines," says Stephen Elliot, an analyst at IDC.
Initially, VMware focused on creating a tool that would allow resources to be dynamically shared across multiple virtual machines on a single physical server, with the eventual goal of managing the entire virtual infrastructure.
"The vision was always to create a highly distributed resource management system where, instead of pegging workloads to specific machines, you could set up a global resource policy for each workload, and the system would enforce that policy across a group of physical machines," says Karthik Rau, VMware's senior director of infrastructure products and solutions.
The first challenge was to develop a tool that could shift workloads among physical servers on the fly.
"When we released VirtualCenter with VMotion, we added one of the key missing pieces to complete the vision: the ability to nondisruptively move workloads between physical machines," Rau says. "With these added capabilities, the team focused on building a true distributed resource manager for the VMware virtual infrastructure, and that turned into VMware DRS."
One hurdle that needed to be cleared was establishing the right algorithms so that resources weren't being consumed moving virtual machines around when it wasn't really necessary, a scenario that would worsen rather than improve performance. The user interface also went through several versions. The product bounced back and forth through a process of design, internal testing and beta testing by 6,000 customers before it was finally ready for a production release.
Using DRS, a manager can group all of the physical servers into a single resource pool and establish policies extending across all of them.
For example, a company's finance department might be allocated a 20% share of the resource pool. If the company has 10 servers running under DRS, the finance department would get the equivalent of two physical servers, though the work would actually be performed by virtual machines running on any of the 10 physical servers.
If a physical server was brought down for maintenance or a new box was added to the server pool, DRS would expand or contract the department's resources accordingly to keep it at 20% of the current resource pool. Or, if a department was releasing a new product, it could be assigned a higher percentage of the server resources during the release period.
A final version of DRS came out in June as part of VMware Infrastructure 3 Enterprise Edition. It can also be purchased at a list price of $2,000 on a price-per-server basis.
Read the original article, here.