Service Provider Uses Virtualization to Boost DR

While most companies have dabbled in virtualizing their physical servers, third-party logistics software and services provider Transplace went the distance by virtualizing 95 percent of the companys servers, storage, and disaster-recovery systems.

You hear it all the time: Virtualization is great for cost savings among non business-critical systems, but it's just not ready to run applications in production.

Whoever still holds those beliefs hasn't had a conversation with Vincent Biddlecombe, CTO at Transplace, which delivers its logistics software to its customers via software-as-a-service, has 500 employees, and about $2.5 billion in annual sales.

A little more than a year ago, Transplace decided that the time had come to move its data center, which then resided on the tenth floor of its corporate office building in Plano, Texas, to a more secure location. During its examination of data center options, the company considered many possibilities, from outsourcing all of its application hosting, to a complete in-house infrastructure refresh. "It became clear to us that we could continue hosting our own applications more cost effectively than outsourcing," says Biddlecombe.

An important part of those cost calculations was virtualization. "By chance, our decision to move our primary data center to a co-location facility coincided with the time we were due for a server hardware refresh," explains Biddlecombe. "That essentially gave us a clean slate to virtualize from the ground up."In eight months time, Transplace successfully moved its data center from Plano to Dallas and converted its IT infrastructure from a physical to a virtual environment -- and the company isn't looking back. In the process, Transplace has saved substantially on software licensing and the number of physical servers it requires; gained the ability to add capacity on demand; revamped its disaster recovery efforts at its Lowell, Ark., disaster recover site; and ultimately enabled its IT team to focus better on key business initiatives.

The on-demand logistics software it delivers to its customers is custom built in Java, running on BEA WebLogic. For its database, Transplace uses Oracle Corp. (Nasdaq: ORCL). Biddlecombe had no trouble converting its underlying Sun Microsystems Inc. (Nasdaq: JAVA) servers, running Solaris, to 64-bit Windows servers running within VMware Inc. (NYSE: VMW)'s ESX Server software for virtualization. For its Oracle database, Biddlecombe moved from Sun hardware to an IBM Corp. (NYSE: IBM) 570 running AIX, which he says flies about twice as fast. The company managed to convert the vast majority of physical servers to virtual machines, but Biddlecombe didn't want to give specific numbers.

To manage storage, Transplace moved all of the virtual servers to NetApp Inc. (Nasdaq: NTAP) storage systems, and used that company's SnapMirror replication software to keep data synchronized between its Plano data center and the Lowell disaster-recovery site. "Everything gets replicated -- our data and our servers -- between both sites. And we have the added benefit of using the DR site for application and environment testing. I don't have 100 physical servers just sitting there collecting dust," says Biddlecombe.

For automated failover between the two sites, Transplace installed VMware's vCenter Site Recovery Manager. "vCenter automates the process of identifying what servers are active in the production data center, and how they need to be mapped to the DR site. It orchestrates the entire failover process."

Moving terabytes of data (just the root directories of each server consumed 30 GBytes each, totaling 3 TBytes), and maintaining full synchronization between both sites utilized an enormous amount of bandwidth and storage. To cut down on storage demands, Biddlecombe ran de-duplication software that helped him to improve storage efficiency by 65 percent. For bandwidth utilization, Biddlecombe turned to Silver Peak Systems Inc. NX appliances, which optimized data throughput between the two data centers.The end result, Biddlecombe says, is the ability to deliver increased application performance and rapid provisioning of new capacity. "If we need more power, we just add another virtualized server. If we need to test an application, we spin-up a new environment." And, should a disaster strike, Transplance can failover to its Lowell data center without skipping a transaction. "Before this transformation, our systems were growing brittle. Now we have more headroom and flexibility. When you spend less time managing aging systems, you have more time to focus on improving the business," says Biddlecombe.