Predicting effect of workload change on system resources

Success of any IT system highly depends upon the performance delivered by the system. Whenever there is a change in workload or workload mix, performance delivered by the system will also change. In this paper, we are proposing a methodology to estimate the consumption of resources at each node of the system for any planned or unplanned change in the workload or workload mix of a web application using the algebraic basis derived from the historic production logs like access-logs and system utilization logs. Our experimental validation reveal that the average error in the estimation for consumption of major system resources namely cpu, memory, disk and network bandwidth was less than 7%. To derive overall resource consumption for the desired workload mix, unlike the traditional approaches, method presented here neither requires to estimate resource consumption for each individual application function nor does it require to benchmark individual business transaction.