Method for determining service demands in a network load balanced scenario
An approach to determine the service demand of various resources of any individual server (present in a cluster of servers), given the overall transaction throughput and utilization values of these resources, in case of a network load balanced scenario is disclosed. The method proposes a quantitative approach to calculate the transaction throughput of each network load balanced server from the overall transaction throughput, obtained from any commercial performance testing tool. The individual transaction throughputs are calculated using monitored values of network related performance counters of each server and the load balancer obtained from performance testing. Service demands of the resource can be computed using the Utilization law of Queuing Theory once the throughput of that server and utilization of that resource is obtained.
1. A method for determining service demands of at least one server from a server cluster having a plurality of servers in a network load balanced scenario, the method comprising:
monitoring network based performance counters during transaction performance testing of the plurality of servers and at least one load generator, wherein performance counters of the at least one server comprise bytes received per second by said at least one server, and said performance counters are monitored using a performance monitoring utility;
measuring percentage of CPU utilization for the at least one server;
monitoring a total transaction throughput of the server cluster, wherein the total transaction throughput is a rate of transactions processed by the server cluster;
calculating transaction throughput of the at least one server, wherein the calculated transaction throughput of the at least one server is a function of (i) the total transaction throughput of the server cluster, (ii) bytes received/sec by said at least one server from the at least one load generator, and (iii) total bytes received/sec by the server cluster from the at least one load generator; and
determining service demand of the at least one server, wherein said service demand is a function of the percentage of CPU utilization measured for said at least one server and the calculated transaction throughput for said at least one server.
2. The method of claim 1 , wherein distribution of overall transaction throughput among the plurality of servers in a clustered environment is obtained.
3. The method of claim 2 , wherein the plurality of servers is a web server or an application server or a database server or any combination of thereof.
4. The method of claim 1 , wherein the calculated transaction throughput of the at least one server is a product of the total transaction throughput and a ratio of (i) the bytes received/sec by said at least one server from the at least one load generator and (ii) the total bytes received/sec by the server cluster from the at least one load generator.
5. The method of claim 1 , wherein the service demand of the at least one server is a ratio of the percentage of CPU utilization measured for said at least one server and the calculated transaction throughput for said at least one server.
6. A computer program product comprising a non-transitory computer usable medium having a computer readable program for determining service demands of at least one server from a server cluster having a plurality of servers in a network load balanced scenario, the program comprising:
monitoring network based performance counters during transaction performance testing of the plurality of servers and at least one load generator, wherein performance counters of the at least one of the server comprise bytes received per second by said at least one server, and said performance counters are monitored using a performance monitoring utility;
measuring percentage of CPU utilization for the at least one server;
monitoring a total transaction throughput of the server cluster, wherein the total transaction throughput is a rate of transactions processed by the server cluster;
calculating transaction throughput of the at least one server, wherein the calculated transaction throughput of the at least one server is a function of (i) the total transaction throughput of the server cluster, (ii) bytes received/sec by said at least one server from the at least one load generator, and (iii) total bytes received/sec by the server cluster from the at least one load generator; and
determining service demand of the at least one server, wherein said service demand is a function of the percentage of CPU utilization measured for said at least one server and the calculated transaction throughput for said at least one server.
7. The computer program of claim 6 , wherein the calculated transaction throughput of the at least one server is a product of the total transaction throughput and a ratio of (i) the bytes received/sec by said at least one server from the at least one load generator and (ii) the total bytes received/sec by the server cluster from the at least one load generator.
8. The computer program product of claim 6 , wherein the service demand of the at least one server is a ratio of the percentage of CPU utilization measured for said at least one server and the calculated transaction throughput for said at least one server.