The concept broadens the reach of virtualization and leverages extra server processing power. Virtualization in the data center is a hot-enough topic, but the concept of server repurposing is extending virtualization in ways that warrant a closer look from VARs.
The notion is simple but powerful, especially for higher-end customers with large collections of servers. Take the parts of a server, such as its IP address, disk storage and associated applications, and make them virtual so they can run on any machine in a data center and aren’t tied to a particular piece of hardware. When an enterprise needs more processing power for particular applications, you can bring up a new server in a matter of minutes and have it satisfy this demand.
This isn’t new: IBM and Hewlett-Packard have been talking about on-demand computing for years. But four things have made server repurposing more affordable to a wider potential audience and more appealing to solution providers. And the companies to watch in this area include Scalent Systems, Cassatt and Egenera.
First, competition has come to the channel. Until recently, server repurposing was primarily a high-end direct play or the province of the largest resellers such as BearingPoint and IBM Global Services. Companies such as Scalent Systems have a two-tier channel program that supports both these larger integrators as well as smaller VARs looking to specialize in this service.
Egenera, which used to only sell hardware, announced last fall it is also selling its software. The company is also “looking for Tier 1 and Tier 2 service providers to sell our product and expand our channels of distribution,” said Susan Davis, vice president of marketing for the company. Egenera’s first OEM partner is Fujitsu Siemens Computers, and the company is also using Apptis and Immix Group to reach the federal government markets.
Second, being green matters more as IT shops try to cut their power and cooling bills, so an increasing number of companies are figuring out how to remotely power up machines and install the virtual machine software on the fly. “We use the existing agents that are already on the machine to do our remote power control,” said Ken Oestreich, Cassatt’s director of product management.
Third, the nature of what constitutes “high availability” is changing. A decade ago, it meant that applications had to automatically fail over without missing a beat, and this primarily applied to transaction-based systems. As enterprises rely more on Internet-based applications, having a few minutes of downtime is more acceptable.
“Now the vast majority of enterprise applications are store-and-forward, and as long as the server comes back online fast enough on a different machine with the same IP address and storage pool, I don’t notice the gap,” said Kevin Epstein, Scalent’s vice president of marketing.
Finally, cluster computing has stalled and become less agile than the virtual machine-based solutions. “The problem with clusters is that they take a tremendous amount of software resources,” said Egenera’s Davis. “The cluster almost has to be custom-built for a particular application and the servers had to be physically identical. The minute you need to change something on that cluster, you are subject to human error.” Server repurposing frees applications from running on particular hardware configurations, so they can run on any machine in the data center.
“Our solution [price] is a fraction of the cost of the clustering software because you don’t have to pay for the additional software licenses,” Epstein said. “At the press of a button, you can rearrange your servers to meet your particular needs.”
David Strom is a technology freelance writer, consultant, blogger and podcaster and can be reached at david@strom.com.
