Measuring selected parameters of a workload in a distributed .NET application by performance counters

摘要

This paper deals with creating and using performance counters, processor (CPU) utilization and memory utilization in a distributed .NET application. It, together with them, uses also the numbers of running processes in computers operating systems of its server and client-side to determine the average values of selected parameters of computer systems workload of both parts. Then a distributed .NET application will use these average values of selected parameters of computer systems workload and also the numbers of logical CPUs in these systems for an automatic division of computational load between its server and client-side. In the experiment, we measure the total execution time of calculations of 10 values of the definite integrals of the selected function f(x) in our distributed .NET application with the automatic division of computational load between its server and client-side and the total execution time of calculations of the same 10 values of definite integrals of the same function f(x), but using the “fixed” division of computational load between these two parts at the ratio, 90% of the load on the server-side and 10% on the client-side. Based on a comparison of these total execution times of calculations as well as comparison of execution times of calculations of particular values of the definite integrals of the function f(x) in these sets of calculations with using automatic and with using the mentioned “fixed” division of computational load between a server and a client-side of a distributed .NET application, we examine the execution effectiveness of the automatic division compared to the “fixed” division of computational load. The results of these comparisons indicate that the automatic division of computational load has better execution effectiveness than the “fixed” division of computational load. Such distributed .NET application, with a functional and effective automatic division of computational load between its server and client-side, can be used as part of the control software of some military equipment.