The Effect of Workload Groupings on Distributed Transaction Capacity Models

Modeling transaction workload behavior with a queuing network model becomes much more difficult when the application consists of multiple workloads. One of the goals of this type of model is to predict how the application will behave when moved to a distributed environment. Add to the situation the goal of predicting the necessary capacity to support the application in the distributed environment and the model quickly becomes very complex. One way to reduce the complexity is to reduce the number of workloads being modeled, but the question then becomes how to combine the transactions into workloads. This paper is a case study of a hypothetical modeling situation where the transactions are grouped differently to determine if such workload groupings have a significant effect on the outcome of a capacity model and whether the effect is the same in a distributed environment. These hypothetical transactions were grouped into several different groups of workloads and each of the groups was studied to see how the groupings effected the response times after the transaction arrival rates were increased. This analysis was done on both a single system scenario and a networked systems scenario.