What was the people side of this transformation? How did you staff these efforts?
Fjeldheim: For the earliest efforts in virtualizing our infrastructure, we created a "tiger team" to lead these efforts. This team helped us develop the initial processes we would use throughout this evolution. Soon thereafter, though, we determined it was best not to separate our virtual teams and our physical teams. That has proven to be a fortunate move.
For instance, instead of having a virtual team that thought only about pushing virtual servers and was competing with a physical team that was pushing physical servers, we had one team that was evolving, based on the natural demand for and evolution of these two areas. From an early stage, this approach also helped us develop one set of management and monitoring tools for the entire infrastructure. That has helped with complexity and with security.
We went from 75 servers per one administrator to 800 [servers] per administrator now. Therefore, what we manage has grown dramatically, but our staff has not grown at the same rate. This frees up our IT employees for innovative activities and means that a larger portion of the IT team is focused on ideas that truly add value to the organization. That tends to energize people and increase job satisfaction on average across the team. Frankly, the growth in the use of various systems also absorbed a lot of [employees] who were no longer needed to work on the physical infrastructure.
After starting with Windows servers, how did you determine what to put into the cloud?
Fjeldheim: The number-one determinant was the usage profile of the infrastructure. Some engineering software can be a real CPU hog. As a result, these instances are not as clearly candidates for virtualization. Those areas that were not CPU hogs were put in the plan for virtualization.
In the first five years, 80 percent of what could be in our private cloud was put in our private cloud. This includes all of our applications running on Linux and Windows servers, but does not include engineering tools [such as] Synopsis and Cadence running on our engineering compute grid, or applications running on big Unix servers [such as] Oracle applications. It took another two to three years to virtualize the last chunk of infrastructure.
Now that it's been nearly a decade since you embarked on this journey, what are the primary advantages to cloud and the SPI model based on your experience?
Fjeldheim: Our first concern is meeting the needs of our business. Demand for IT services always seems to be increasing, and the patience of the business is decreasing. This presses IT to operate with agility. My colleagues outside of IT are some of the savviest technologists around. If I can't meet their needs, they'll find other ways to be sure their needs are met. I need to be sure that we are always the first choice when it comes to technology development. We have been in growth mode as a business for a while, and I'm proud of how IT has supported that growth. The more often we can provide value-enhancing technology, whether it is new software or the infrastructure backbone, the better we will serve the needs of our customers ultimately. Cloud computing and the SPI model are a big part of allowing us to get there.
We have found that the biggest benefit for us was speed. We can deploy a virtual server in less than 15 minutes, where it used to take up to six weeks to put a new physical server in place. It used to take days to get a server up and running, and it would come out of the other departments' budgets. Now it takes five minutes, and the cost is in my budget. Needless to say, costs are dramatically lower.
The cost savings of this approach can't be overstated. We believe we have generated $25 million to $30 million in savings in hardware and power. We have also significantly improved our service levels, as unplanned downtime dropped by almost 90 percent when going from physical to virtual servers. Moreover, this allows us to render variable a lot of what were once fixed costs.
We can ramp up our infrastructure, for instance, at the same rate that demand for the different facets that make up our infrastructure is growing, rather than having to purchase significantly higher capacity than what we need. For other companies that are unfortunately in cost-cutting mode, [the lesson is that] this also allows them to make cuts in infrastructure before having to cut the firm's most valuable resource: its people.
What advice would you offer to CIOs who are at the beginning of the journey on which you are now so far along? If a CIO is considering the cloud and the SPI model, how would you counsel him or her to plan and enact this transformation?
Fjeldheim: My first piece of advice would be to "go big" on these changes because we have found the deeper we got into cloud computing and the SPI model, the more value was created. I know a lot of CIOs want to pilot this and bite off the various aspects in very small chunks. I know that many of my peers have had an experience similar to mine when they pushed aggressively for these changes.
Second, plan for success. Think about the people and processes to guide the process, and ensure that you can ramp up quickly.
Another people point is to help IT and the rest of the organization understand the value they can get from this sort of transformation. These days, companies have many examples of successful transitions such as the one I have described here. This is an advantage I would have liked to have had when we embarked on this journey.
It is also important to think of the human element in this equation. Retrain your staff so they can adequately support the growth in cloud computing and the SPI model.