Rants
Questions
Soapbox
Best Practices
Apply today for a FREE subscription to CIO Magazine!
Tue, Sep 9, 2008 13:59 EDT

|
Posted by: Bernard Golden in Best Practices Topic: InfrastructureBlog: The Open Source
Current Rating: |
Gartner echoes my posting of Friday on Cloud Virtualization. Except they haven't got it quite right -- they don't quite reach the logical implications of their statements.
In the article reporting Tom Bittman's remarks at a conference, (he is Gartner VP and Chief of Research of Infrastructure and Operations), it recounts his description of cloud computing as being based on four elements: "services-orientation, utility pricing ("whether it's subsidized, or I pay for it based on use"), elasticity ("perhaps massive elasticity"), and that it is delivered over the Internet." The first three of this list, Bittman says, are also present as trends within data centers. Therefore, cloud computing and data center virtualization share similar factors.
He is completely right about these elements being common to virtualization and cloud computing. He just doesn't go far enough to tease out the logical extension of this: if they share common elements, there's really no difference -- they'll end up being the same phenomenon, just applied within different use cases, which will be decided at run-time, not design time. What does that mean?
It reflects what I said in my last posting -- that the boundary between data center and cloud is going to get porous. Businesses will begin to view in-data center and in-cloud as merely two pool of resources, selected according to criteria set at the time of system turnup. In other words, applications will be designed assuming a basis of virtualized resources; when it comes time to deploy them, the decision about whether to host locally or in the cloud will be made on the basis of external (external to application design, that is) requirements: network bandwidth required, any relevant security conditions, resource availability. Some applications that require very high bandwidth will be located locally to ensure best possible responsiveness. Others that are less network-intensive, but more CPU-intensive, will be put into the cloud, where processing power is available in an on-demand fashion. And applications may be shifted back and forth between the data center and the cloud as conditions change. Applications may even execute in both regions simultaneously. If the application is architected in a virtualized fashion, the locus of execution is irrelevant.
So Gartner accurately points out that silo architectures are passe, but still falls into the mental trap of defining the data center and cloud as separate execution silos. It reflects the existing mindset of "I own what's in the data center, and what's out there on the Internet is something very different from what I own." In the past, that was true. It's still true to a large extent. But given where infrastructure is going to, it won't be true in the very near future.
Bernard makes a good point: the solid walls that make up today's data centers are starting to look more and more just like cube partitions. Beyond the walls stretch miles and miles of virtual servers that stand waiting to be used.
Jud to drive the importance of this unlimited standby capability home, just think about the outage that the London Stock Exchange (LSE) experience yesterday. Their trading volume grew too quickly and caused their software/hardware capabilities to be exceeded. I almost hate to use the term, but could "cloud computing" be the solution for the LSE? Specifically, should they design their apps to run on their servers in their data center but build in an option to expand to additional servers located in some secure cloud in the event that there is a surge in trading like (tried to) happened yesterday? You can never guess at exactly how much computing capacity that you'll need and perhaps this is where the brave new world of cloud computing can shine. For the next LSE CIO...
- Dr. Jim Anderson
Blue Elephant Consulting - The Accidental IT Leader Blog