Neither cloud computing nor virtual servers were intended as agents that would change traditional IT organizations, says Rachel Dines, a researcher at Forrester Research who specializes in IT infrastructure and management. But IT organization and management issues are turning out to be nearly as important as the technology itself to making large-scale virtual-server migrations effective.
By making it possible to make granular IT resources like CPU cycles, memory, storage capacity and I/O bandwidth available to machines or software far beyond their physical space, virtualization technologies forced IT organizations to break down management silos built on those physical limitations, Dines says.
"If you think about it, it kind of makes sense to have all the system administration for a certain data center in one group defined by physical area," she said. "The first thing that happened with virtual servers is that the VM administrators had to reach beyond that geographic space and get involved with the applications and the business units they were working with."
If virtualization had stopped with a few servers, the pressure to reorganize would have stopped after enough sysadmins had learned to manage both virtual servers and how to work for more than one department at a time, says James Staten, principal analyst at Forrester.
But the virtual world is a lot more dynamic than the physical one, and it crosses a lot of traditional organizational boundaries, notes Chris Wolf, research vice president at Gartner.
"Without rethinking some of that organizational structure and aligning at least some portions of itself more with the virtual infrastructure, it wouldn't be able to keep up," Wolf says.
The ROI and increase in technical efficiency is so high in the first rush of virtual-server migrations, it makes up for a lot of gaps in support and organization, Dines says. But long before say half of a company's servers can be virtualized, organizational weaknesses can slow the migration and limit the success of existing migrations.
Though virtualization and cloud are both too immature for IT to have smoothed out all the organizational wrinkles, some consistent management principles have emerged that can help you avoid trouble.
1. Widen your focus:
It seems simplistic, but shifting a server from physical to virtual eliminates its allegiance to the physical box on which it lives, Wolf says. That means sysadmins can change their focus from the boxes in their immediate area to all the virtual servers handling load balancing, or security duties, or all the VMs running applications for a particular business unit.
"You can get efficiencies that weren't accessible before organizing by function," Wolf says.
Administratively that could shatter the relationship between workers and managers in the same facility, and reassign them all along functional lines. A small number would remain tied to the physical plant and physical servers, but most would be available to be reassigned to the highest-need functional areas.
"Making that work can be a real challenge," for traditional IT organizations, however, Wolf says.
2. Flatten the hierarchy, widen the job descriptions
The same dynamic that assigns IT workers to the functions that need them most eliminates the need for several layers of management between the CIO and first-level managers, Dines says.
Regional managers whose role is to coordinate activity within a geographic area could be reassigned to more innovative project-development work or more sophisticated capacity planning and operations management, Dines says.
"In the virtual world you don't have the equation where you add another box when you need capacity. Adding capacity is easy, you just spin off another VM," she says. "Real capacity planning in a virtual environment means planning for the whole data center, not just one set of applications or data center. It's a much different set of skills than it used to be."
It's also so critical to the success of both the virtualization project and overall IT performance that it is becoming a separate department or set of job descriptions for people who can take into account not only the capacity of computing hardware, but also anticipate needs of business units.
"Covering spikes is a pretty basic form of planning; this is more approaching capacity as an optimization process," Dines says.
3. Respond like a service, not a utility
IT's traditional role has been more support than enhancement, according to Rob Smoot, director of product marketing for VMware's vCenter management products.
As virtualization expands and business-unit managers get more used to the idea they can order up high-level IT services from external cloud providers "and pay for them with a credit card, not ask IT for budget to do it," their relationship with IT changes as well, Smoot says.
"Instead of just dealing with one application for a business unit, IT is responding more to a wider range of business needs. As that happens IT has to become more of a service that supports the business and allows users in the business unit to do more self service, as they become more comfortable with app stores and provisioning their own servers," Smoot says.
That relationship is both more and less work for IT -- because it requires a different set of rules and processes to allow end users to do more things for themselves, but still protect data that needs to be protected and not allow self-service to expand into capacity eating sprawl.
"You need a more real-time view of the environment and the ability to respond to it with mature process, not do it in an ad hoc way," he says.
4. Train, don't hire
All those organizational changes and new job requirements usually require a lot of new hires people with skills not present in middle management or the rank and file, according to John Reed, executive director at Robert Half Technology, the IT recruiting arm of Robert Half International.
"We're seeing a lot more job descriptions with cloud or virtualization in the description, and some change in descriptions. A lot of companies are saying the requirements are complex enough and knowledge of the company are so important that they prefer to train up people inside for the critical jobs," Reed says.
The most critical have the word "architect" in the title, and not just for the top job.
In virtual and cloud implementations, architecture is a combination of capacity planning, design and implementation, workload distribution and functional knowledge of the workloads being run on those virtual servers.
The need for bodies to do the work is too great to meet it only with retraining, however.
"The number of jobs posted for jobs with cloud computing or virtualization is growing daily, growing exponentially," Reed says. "Even job descriptions that aren't primarily cloud related networking or security analysts typically have some responsibility for handling those functions as they relate to the cloud, protecting data as it moves back and forth, or working with an external app vendor to set layers of security for specific apps in specific circumstances.
"We're definitely seeing a major transition in the marketplace," Reed says.
5. Fight to Automate
Process maturity and automation aren't exactly he same thing, but they overlap a lot, according to Frank Gillett, vice president and principal analyst at Forrester.
Most companies aren't ready to make large-scale organizational changes to accommodate virtualization, even if they really believe in the technology, Gillett says.
Critical as the organizational issues are, automating security, provisioning policies and all the other disparate demands of virtualized infrastructures takes a lot of technology as well.
"Trying to get to cloud, there's a whole implementation layer self-service functions, portals, resource-consumption trackers, chargeback to users that's based on consumption rather that the number of servers there is a lot of automation technology required to manage the whole thing effectively," Gillett says.