It's easy to believe that a system's utilization increases in a linear fashion. A system that "grows" by 10% annually will be utilized 30% for year 1, 40% for year 2, and 50% in year 3.

There are several problems with this sort of capacity management:

• A lot can happen in a year, quarter, month, or day. Does this account for variance in the utilization, or is it just an average?
• Averages can be dangerous also.
  • A CPU/Network/whatever that was 100% utilized for 8 hours yesterday, and completely idle the rest of the day averages out to a 33.333% utilization for the (24 hour) day.
• Today's operating systems handle higher CPU utilization much better than before, the scheduling algorithms used reduce the complexity of process/task scheduling.
  • While this reduces the amount of "thinking" the operating system has to do, it does not reduce the amount of "heavy lifting" it has to do to accomodate the increased load.

Drawing a line through a series of points on a graph is not a reliable method of forecasting capacity requirements - Especially as a system's utilization gets closer to being saturated because of ancillary workloads and resource availability. An example of this is a situation where a system's CPU is capable of generating more I/O than a storage system is capable of handling.

We identify and quantify the limits, variance, trends, scope, and scalability of your systems. This information is imperitative for capacity management.