While many IT executives mull over the notion of buying infrastructure as a service under the new cloud computing model, Chad Swartz, Senior Manager of IT Operations at Preferred Hotel Group, is actually doing so. Last December, Swartz powered down the servers and storage at this Chicago-based premiere hotelier and started tapping into capacity he buys from cloud provider Terremark Worldwide. In this interview with ITworld contributing writer Beth Schultz, Swartz reflects on ...

While many IT executives mull over the notion of buying infrastructure as a service under the new cloud computing model, Chad Swartz, Senior Manager of IT Operations at Preferred Hotel Group, is actually doing so. Last December, Swartz powered down the servers and storage at this Chicago-based premiere hotelier and started tapping into capacity he buys from cloud provider Terremark Worldwide. In this interview with ITworld contributing writer Beth Schultz, Swartz reflects on ...

BUYING INFRASTRUCTURE FROM A CLOUD PROVIDER: Giving up so much control over your environment is more difficult than I thought it would be. Being reliant on other people to recognize and fix your problems is definitely a hard thing to do.

TRUSTING YOUR CLOUD PROVIDER: You expect the cloud support staff to know what' + char(39)+ N' + char(39)+ N' + char(39)+ N's going on with your infrastructure, to diagnose problems and to get your systems up and running as soon as possible. But you always have that little conversation going on in your head, "If I just had this in house, I would have had that problem fixed two hours ago." That' + char(39)+ N' + char(39)+ N' + char(39)+ N's a bit frustrating.

THE IRONY OF CLOUD COMPUTING: We just started hosting our Web sites in house whereas before we outsourced them. So we' + char(39)+ N' + char(39)+ N' + char(39)+ N're learning that, because of the flexible nature of bursting and stuff like that, we can take on more and better projects. And we can really save a lot of money that way. Now that we have the infrastructure in place, with load balancing and such, bringing outsourced stuff inside becomes viable, which is sort of ironic - outsource the infrastructure so you do not have to outsource other things. But it' + char(39)+ N' + char(39)+ N' + char(39)+ N's all about consolidating your operations.

TURNING BACK: There is no turning back. Once you get past a certain point of buy-in, you' + char(39)+ N' + char(39)+ N' + char(39)+ N're on the path. Granted, I could do a hybrid solution where we could have some of their stuff and some of ours and link them together, but I think that' + char(39)+ N' + char(39)+ N' + char(39)+ N's counterproductive as far as cost-savings go. You still have to do everything you did before, but now you' + char(39)+ N' + char(39)+ N' + char(39)+ N'd have an extra link in there. You would lose a lot of the gains you have by having infrastructure in the cloud by going to a hybrid solution.

WHAT TO PUT IN THE CLOUD: If I had known differently, I would have built servers from scratch instead of doing P-to-V conversion. There are a ton of things that had I known starting from Day 1 could have had made for an easier transition, but that would be a primary one. Things cannot operate the same way.

Besides that, I would have invested more time in data clean-up, with the obvious reason being to have reduced the migration time required. I also would have used the time to replace or upgrade software versions.

I know that a lot of people would be hesitant to replace too many variables during a transition. However, if you' + char(39)+ N' + char(39)+ N' + char(39)+ N're building servers from scratch it' + char(39)+ N' + char(39)+ N' + char(39)+ N's better to get it done one time. In our scenario, we had to build a second time to accommodate upgrades to our Citrix servers and the applications we use.

HOW TO ALLOCATE RESOURCES IN THE CLOUD: We' + char(39)+ N' + char(39)+ N' + char(39)+ N've learned it' + char(39)+ N' + char(39)+ N' + char(39)+ N's best to under-allocate a server at first rather than over-allocate. You just want to have resources run efficiently without too much excess sitting out there because that takes away from the effectiveness of other servers being able to grow dynamically.