Arthur Cole spoke with John Engates, CTO, Rackspace.
Cole: You claim to offer a private cloud based in your own data center. How does that work?
Engates: The private cloud is the message you hear from companies like VMware, EMC and Cisco. They spend a lot of time talking about enterprises building their own cloud, and VMware's VSphere and other new stuff in the works is the way to do that.
I think it's interesting, though, that people have built their own cloud in their own data center, even though it defeats the purpose of the cloud in that they are not taking advantage of others' facilities, hardware and management. And the pay-as-you-go model goes out the window.
A private cloud in our data center, with our hardware and infrastructure, gives back the utility aspect that you would lose if you built it yourself. It's how enterprises can tap into the power and technology of the cloud and reap all the benefits it has to offer.
Cole: But how can it be the customer's private cloud if you own all the infrastructure?
Engates: Because we dedicate a firewall to the customer in our data center. Private clouds are always done on dedicated infrastructure behind the private firewall. Ours is the same because it is still behind the customer firewall that we have dedicated to them in our data center. It's no different from a customer having a co-located data center because we are not sharing any hardware resources at all. You have a dedicated firewall, a dedicated network switch, load balancing ... all are dedicated. It's your own remote infrastructure.
Cole: Won't the customer give up some measure of management or control by outsourcing the cloud to you, particularly those that need to scale up or down on a moment's notice?
Engates: We're using standard VMware 3.5 infrastructure, which gives you the ability to turn virtual machines on and off, to put them in standby or hibernate mode. Customers don't pay for the licenses when the machines are turned off. You pay for the underlying hardware infrastructure, but the licenses and fees all go into a suspended state.
What you trade out is some of that instant, infinite scalability that you get with an internal cloud. But we think most people who want our cloud are not the ones who go from one to 100 machines and back again. You can deploy quickly when you need to, but it's not really designed to be used in the fashion of someone on EC2 who scales up and down constantly. Enterprises typically have well-defined and constant needs. They are not as bursty as Web-centric or 2.0 companies.