Alistair Croll is "interviewing" Lew Moorman, CSO, of Rackspace. I've captured some, but not all, of the conversation.
AC: From your recent filing, you have 62,078 servers. 69% are cloud. Rest are managed. What's the difference?
LM: Cloud is set of technology about pooling and automated software. Cloud makes shared environments robust and reliable. Previously, in a shared environment, if someone else burst, everyone else loses.
Lew describes dedicated hosting as a house, and cloud as a hotel room.
AC: You have one set of costs associated with virtualized servers and another for dedicated. Where do you (financially) want customers to be?
LM: Don't break out the economics. Cloud is more capital efficient, but very competitive which impacts margins. In dedicated environment, customers ask for specific server and configuration. Harder to redeploy when customer leaves.
AC: Why do customers ask for specific configurations? Is it server hugging?
LM: Might be for security, for cutting edge technology implementations, for performance, or just preference.
AC: What's the biggest driver for your customers?
LM: Rationalizing your computing operations. Economic climate, customers need to save money.
AC: Will you launch a sharded service, similar to your email service?
LM: Dedicated to Cassandra project. Excited about Drizzle. Needs to be a next generation datastore that people use. Needs to be standards based. Won't commit to service offering. But, data scaling without huge operational overhead is a problem customers need solved.
... Using these new data stores and techniques is a commitment. Need to change applications. Service providers need to make possibilities available and help educate. The final technology choice is up to the company.
... We are not an integrator. There will be huge market in professional services to help people scale, efficiently. Easy to scale, not always efficient. This is not a market for Rackspace.
AC: Seems like engineering needs to tie into user experience more, with a conversation on accounting. "X performance adds n cost. Should we do this?"
LM: Dealing with IT sprawl is first order problem for enterprise IT. Scale is less common driver.
AC: Mosso still based on VMWare?
LM: No. Didn't meet the needs of a service provider at massive scale. Prefer to leverage existing, but will build solutions as required - feature/function and economics.
... Using Xen on Linux.
AC: Does this make it problematic for companies to move internal VMWare image to your cloud?
LM: There will be machine image standards in the future. There are tools to move images between clouds. People do move between Rackspace and Amazon today.
AC: How can you position (and survive) on 'fanatical support' as a utility company?
LM: Cloud computing is not going to work as perfectly as power out of outlet. Lew calls out attendee poll yesterday, no one raised hand to "will cloud computing run perfectly". Also ties to Jesse Robbin's lightening round talk on Failure Happens.
People will want service to assist in cloud setup, run and evolution.
AC: Does cloud need reference model for assessing true cost of ownership?
LM: Glad McKinsey did study, sparked a lot of conversation. The biggest flaw was it assumed none of benefits of cloud computing, enterprises need excess capacity.
AC: Final advice for IT?
LM: Don't listen to sales folks, like me. Go use it! Go experiment. Solve problems that exist in your company.