2008 InfoWorld CTO 25: Vincent Biddlecombe, Transplace

A combined datacenter refresh and SaaS platform project provide an opportunity to change the status quo

 

As CTO of logistics services provider Transplace, CTO Vincent Biddlecombe tackles the usual projects such as implementing disaster recovery plans and selecting vendors for tools such as business intelligence. But his particular talent is thinking about how to use internal IT to serve customers. Part of that is being a logistics services provider -- much of the product is in fact driven by IT, which is why nearly a fifth of the 500-plus employees at Transplace work for Biddlecombe. But part is seeing infrastructure as more than just infrastructure.

Case in point: Like many companies, Transplace is applying virtualization technologies to its datacenter infrastructure to improve server utilization and facilitate disaster recovery and resource flexibility. And as in some other companies, Biddlecombe is exploring providing Transplace's services to customers over the Web as a SaaS (software as a service) offering. But Biddlecombe recognized that both initiatives are related.

[ Discover what insights you can take advantage of from the other 2008 InfoWorld CTO 25 winners. ]

"The two are hand-in-glove," he says. "If we're refreshing the hardware, let's do a world-class infrastructure," he says, that can handle both internal IT needs as well as provision customers via SaaS.

Begun a year ago, this effort required Biddlecombe to stretch himself and his team. For example, "historically, Transplace is a Sun shop -- all Java-based -- with an Oracle database and Solaris servers," he says. But after exploring his options, Biddlecombe came to realize he needed a different platform, despite Sun's great fit in the past.

The initial issue was performance: To provision via SaaS, Transplace needed speedier, more capable systems. In the Sun world, that would have meant more servers, and thus more Oracle licenses -- increasing both management complexity and cost. By shifting to servers running IBM's Power6 CPUs, which run twice as fast as Sun's, Biddlecombe could gain that extra capacity without requiring more licenses or filling his datacenter with servers. So he switched to an IBM 570 and AIX for his database, 64-bit Windows XP on EMC VMware virtual machines for the middle tier, and NetApp for storage. "The easy path would have been to keep all Sun. But we had an opportunity for a major change, so we need to think outside that comfort level," he says.

Not only did Biddlecombe take the risky step of switching out a platform with which his team was very familiar, he also pushed the use of virtualization. It quickly became clear that virtualization is not a simple technology to adopt. "There is no holistic, prescriptive architecture for virtualizing," he notes.

Most VMware implementations are for app servers, but he was also using it for database transactions. He had lots of questions he couldn't get clear answers to: How many LUNs should you have? Which drives should you use? How many VMs can you place on a LUN? Should you replicate the swap file? What are the right implementation approaches for an Exchange server, for a database server, and for a transaction server in a virtualized environment?

[ For all the latest developments in virtualization, see InfoWorld's Virtualization Topic Center ]

The virtualization and storage vendors were of little help, even when Biddlecombe brought them together to answer his questions jointly. "I asked them, 'Do each of you have an idea on how to do it better?' But their view is narrow to just their product."

So his team had to figure out the right approaches largely on their own. "We've come up with our own models, but we know we can do better," Biddlecombe says. One lesson: "We had to split the VMware environment because it wasn't duplicated for disaster recovery." That means each app is replicated, and "if one environment fails, we still have the other."

To ensure that the new virtualized datacenter would deliver on promised service levels to SaaS customers, as well as to Transplace itself, Biddlecombe instrumented every virtual machine and application, as well as the Java queue, at a low level to get an early warning of any issues that could hurt service delivery. This is hard to do with packaged apps, but that wasn't an issue for Biddlecombe: "Everything we write is custom code, so we understand where to put the probes."

Biddlecombe's combined datacenter refresh and SaaS delivery effort underscored his approach to managing IT: "You can't just accept the status quo."

Galen Gruman is executive editor of InfoWorld.