Why does it matter if we don’t verify initial performance of new hardware like a CPU?
CPU models can vary widely in how they handle a certain mix of CICS and batch in a workload. While measuring initial performance of a new CPU against your actual workloads might seem daunting, ignoring this step can mean a significant increase in unexpected costs.
For example, if a midsize shop with 1500 to 2500 MIPS sees a 4 percent increase in CPU usage with its actual workloads on a new CPU, it could end up paying $60,000 to $120,000 more per year in IBM software alone. The problem is even more costly for larger shops. If you can determine that the extra CPU usage is attributable to how the new CPU handles your actual workload mix instead of due to latent demand, you can potentially renegotiate your contract to lower your costs.
Also, if you calculate chargeback, it’s important to set a fair rate to your internal customers, or agree on a fair rate with your outsourcers. CPU validation is a crucial part of that process.