Loading ...They say you always remember the name of your first love and your first customer. I know I remember my first customer very well.
This was the mid 80s. Mainframes still ruled, and ran most of the world. We were selling a financial services company a new mainframe to help them modernize their IT department.
It was also a competitive win-back. While the CIO (or Data Center Manager as they called him), was clearly a key player, so were two other people — for completely different reasons. The first was the Head of IT Strategy, what today we might call the CTO, and the second was the Head of Capacity Planning.
The Head of IT Strategy at large companies often lived at what one of my colleagues used to call the “reading edge of technology.” They understood all the latest thinking and all the latest technology that had yet to be deployed. They said things like “in the next few years it will all be about objects and RISC.” RISC yes, objects, not so much.
The Head of Capacity planning was a much more interesting guy. Not covered in glory but at the core of the decision cycle. It was he who decided how much capacity was needed and, more importantly, when it was needed. A good guy to be close to – especially if you sold the capacity!
In the 80s, capacity planning on a mainframe was pretty easy. Each different type of user needed a different per cent of MIPS (MIPS was a measure of power that was used to rate the machines). Add them all up, decide what the peak usage was, what level of utilization you found acceptable (say 85%) and you were on the way to knowing what machine to buy and when.
Now jump forward 25 years. How much has changed?
We still look at workloads but they all sound very different.
Today, IDC defines workloads as the “what” folks are using their server for (CRM, ERP, Web serving, etc). Environments like virtualization and operating systems are the “how” folks are implementing and hosting their workloads. That being said, if you extract the virtualized component for each of the major workload categories, servers purchased to host virtual machines have been the largest growth driver and made up about 12% of server sales in 2007*. In fact, more than half of all spending on virtual servers is in support of business processing and decision support applications. Non-virtualized components of Web Infrastructure, File & Print, and Email made up the other majority of workloads.
Today, IDC defines workloads as the “what” folks are using their server for (CRM, ERP, Web serving, etc). Environments like virtualization and operating systems are the “how” folks are implementing and hosting their workloads. That being said, if you extract the virtualized component for each of the major workload categories, servers purchased to host virtual machines have been the largest growth driver and made up about 12% of server sales in 2007*. In fact, more than half of all spending on virtual servers is in support of business processing and decision support applications. Non-virtualized components of Web Infrastructure, File & Print, and Email made up the other majority of workloads.
While we had databases in the 80s – they were not the largest workloads (repeat after me “VSAM”) and we would not have seen Web serving or File & Print as separate activities. Interestingly, by the late 80s and early 90s, virtualization began to appear as a usage model.
Like the planners of 25 years ago, today’s x86 capacity planners have the same challenge, but a much lower utilization rate to work on. While average x86 machines may run at sub-20% utilization, with virtualization that might double or even triple. Web workloads are completely different from database workloads and each has to be measured in, and outside of, a virtualization environment.
Also, like planners of 25 years ago, one benchmark does not tell the whole story. You need a basket. The capacity planner in 2008 needs to look at all the numbers and then plan for growth.
Now, you might be saying “I wonder if he’s writing this because their benchmarks don’t win.” Actually, in the key areas of Virtualization, Web and Database, we are currently the world leaders. But that’s because we have a better architecture – maybe a story for a different day.
So, for capacity planning purposes only ….
*IDC Multi Client Study, Workloads 2008: Understanding Server and Storage System Deployment.
Nigel Dessau is senior vice president and chief marketing officer at AMD. His postings are his own opinions and may not represent AMD’s positions, strategies or opinions. Links to third party sites are provided for convenience and unless explicitly stated, AMD is not responsible for the contents of such links sites and no endorsement is implied.
(4.69 out of 5)