I wrote not long ago about the various disciplines that data center operations teams will need to work through to address those cloud-computing values you often hear hyped by people like me.
In that post, I noted that many organizations had gained an understanding of how server virtualization could be used to abstract software concepts, thus managing them distinctly from the underlying hardware. I also noted, however, that few organizations had made the decision to systematically automate that management.
Channel-V tonight pointed me to an interview by Virtualization Review's Keith Ward of Bogomil Balkansky, VMware's senior director of product marketing. In the interview, Balkansky discusses the upcoming VDC-OS product release, and what it means to the next generation of data centers. He starts with a very familiar theme:
"Henry Ford introduced automation to the manufacturing world," Balkansky says.
"We're transitioning from swinging hammers to pushing buttons," he continues. "The focus becomes on what needs to happen, not spending the majority of your time executing it and making it happen. Ford introduced speed and efficiency and predictability in the (manufacturing) process." Those same elements will characterize VDC-OS, he says.
Balkansky goes on to point out that the very core of the system administrator role will change as a result, an argument that I've been making for some time. Rather than focusing on reactive, tactical operations, the system administrator of the future will "specify the service levels the application requires: availability, security, scalability."
"(Whether an app) needs five-nines availability, security--it needs to scale to this point. Then the infrastructure that supports and runs the application will interpret policies, execute and guarantee service levels, and do it at the lowest total cost of ownership possible. TCO is really a parameter" in VDC-OS, Balkansky adds.
(I pause here to note that he refers to the lowest TCO for the service-level policies that you have in place. You can still run an expensive application profile if you enforce five-nines, bullet-proof security and sub-second response times.)
Channel-V points out that Citrix is also on the case:
Citrix is also working on a similar solution with Citrix Workflow Studio, which is based on a power shell so it opens up to other solutions from other vendors...It is all about automation, but what about automation in cross-vendor data centers? How will that work?
It will work through accepted standards and protocols between components at all levels of the distributed compute stack. VDC-OS and Citrix Workflow Studio are pieces of the puzzle, but there is much work under way throughout the IT industry.
What does it all mean? Well, if you are using one of these two virtualization platforms--and I would guess that Microsoft isn't far behind here--you can bet that the move from manually managing VMs on a tactical basis to strategically automating VM management will be made much, much easier as 2009 progresses.
If you don't have an automation pilot in your budget for this fiscal year, I would seriously recommend planning one for the next cycle. I would also strongly recommend that system administrators begin to think about how they would automate their jobs. Better yet, how would you eliminate the need for the human-centric processes you use to manage VMs altogether, in favor of tight metrics-centric optimization policies? Focus on being a strategic thinker this year, rather than being bucketed as a tactical "resource."
Any way you cut it, the move to virtual machines has opened the door for policy-based automation (aka service-level automation, or SLAuto) and its time to get ready to
rumble automate. Once you do, self-service and metering--the heart of cloud computing--are just around the corner.