How to become an accidental cloud service provider

Eric Knorr| Infoworld

It’s always tough to generalize about which customers are betting heavily on the cloud and why, because the use cases are all over the place. I made an attempt recently to nail things down in an interview with Scott Sanchez, director of strategy for Rackspace Open Cloud, asking him to admit that the real target for the OpenStack cloud platform was the service provider market.

Sanchez’s pushback was enlightening: “You’re right that our target market is service providers. But what we’re seeing is large enterprises that want to become service providers.”

Sanchez meant that in a broad sense — large IT departments are demanding the cloud automation necessary to deliver IT as a service, mainly to internal customers. But I think many companies that make a major commitment to either the private or public cloud end up becoming cloud service providers to a variety of customers, whether or not they planned it that way.

A great example of that transformation emerged from a conversation I had with Van Beach, product manager at Milliman, one of the largest independent actuarial and consulting firms with 55 offices around the world. I interviewed Beach in a cloud panel discussion at the Microsoft U.S. CIO Summit on the Microsoft campus last week.

“If you talked to me four years ago, I suspect I would not have envisioned myself sitting on stage today saying I’m a cloud service provider,” Beach told me. “But that’s really what we are and I think we’re doing pretty well at it.”

Getting in on the ground floor
Beach’s journey began in 2010 when Milliman worked with Microsoft to move its flagship application, MG-ALFA, to the Windows Azure public cloud — the same year Azure officially launched. Originally offered as licensed software, MG-ALFA is a highly processing-intensive application that insurance firms typically run on a grid infrastructure for risk management analysis, employing simulations with randomly generated economic scenarios.

The initial attraction of the public cloud was the spiky nature of the workload. “At year-end or quarter-end you’re dealing with high levels of demand, and off-peak it’s very low,” said Beach. “So to meet that challenge, you either have to … balance those workloads more evenly across the entire year, or you buy resources for peak demand and you’re dealing with low levels of utilization and ultimately have a very costly solution.” Such variability is often considered an argument for turning to the public cloud, which features both near-infinite scalability and pay-per-use pricing.