April 14, 2010, 7:58 AM — by James E. Gaskin - This tip comes from Charles Babcock, Editor at Large for InformationWeek magazine and author of Management Strategies for the Cloud Revolution, published by McGraw-Hill. Babcock wrote multiple stories about various aspects of cloud computing and decided a longer form than magazine articles was needed. 272 pages later, he had the book.
1. First off, Babcock defined "cloud computing" because every vendor applies a different slant. Cloud computing is a new method of distributing computing power. All the pieces have been around, but the combination of high bandwidth availability and relatively low cost cluster servers make cloud computing practical.
2. Cloud computing will help you get out of the bad habit of over-provisioning your own data center. Need an app? Get a server. Need another app? Another server. That's how we reached the point where data centers are full, but averaging only between 8-15 percent of capacity. Sure, there are exceptions to the average, which brings us back to cloud computing as a way to handle that overflow without wasting capacity on your servers every other day of the year.
3. Some data will always stay private, period. No matter how good cloud security becomes, some companies will always be wary, and some corporate data must be tracked constantly for compliance issues.
4. This could be #3a, because cloud computing vendors can't always tell you where your data is, exactly. Data and workload replication provides excellent disaster recovery support and the horsepower needed to handle transaction surges, but that means the data center you visited and vetted may no longer be the location of your data.
5. Know your workload before searching for a cloud computing partner. Your partner selection sheet should include details about your workload including CPU types, average workload and capacity, and whether the workload is storage or transaction intensive. HIPAA or PCI data? You must know, and your cloud partner must know, and you must agree on how they should handle (or not) your regulated data.
6. You will have a mix of public and private clouds. Corporate data centers, full of x86 servers with virtualization in progress, are most of the way to a private cloud already. Create server templates for your workload, as many variations as necessary, and talk with your cloud computing partner about replicating those in their data center.
7. Match your virtualization hypervisor to a cloud computing vendor using the same type. Shop for a partner who can accept your templates, making it simple to roll your servers out into their cloud support structure.