CPU Utilization Micro-Benchmarking for RealTime Workload Modeling

The major challenge when migrating platform to satisfy the current and future computing demands is to decide which is the most optimal option for migration. Without actually executing the workload in a platform, it is difficult to know the workload performance in the platform. However, comparing the workload performances between different platforms by testing the workload is very tedious and time consuming. This motivates us to design a modeling framework to predict the workload performance of CPU on different platforms without executing. The challenge for the modeling lies within the collection of highly correlated data to train a predictive model. In this paper, we present a novel CPU utilization (%CPU) micro-benchmarking method to collect the data needed as a vital step before proceeding to training phase.