Predicting Performance Regression of a Computer System with a Complex Queuing Network Model

摘要

An approach is provided for predicting system performance. The approach predicts system performance by identifying a Queuing Network Model (QNM) corresponding to a clustered system that handles a plurality of service demands using a plurality of parallel server nodes that process a workload for a quantity of users. A workload description is received that includes server demand data. Performance of the clustered system is predicted by transforming the QNM to a linear model by serializing the parallel services as sequential services, identifying transaction groups corresponding to each of the server nodes, and distributing the workload among the transaction groups across the plurality of nodes. The approach further solves analytically the linear model with the result being a predicted resource utilization (RU) and a predicted response time (RT).

主权项

1. A method, in an information handling system comprising one or more processors and a memory, of predicting system performance, the method comprising:
identifying a Queuing Network Model (QNM) corresponding to clustered system that handles a plurality of service demands using a plurality of parallel server nodes that process a workload for a quantity of users; receiving a workload description that includes server demand data; predicting a performance of the clustered system, wherein the predicting further comprises:
transforming the QNM to a linear model by serializing the parallel services as sequential services;identifying transaction groups corresponding to each of the server nodes; anddistributing the workload among the transaction groups across the plurality of nodes; and analytically solving the linear model, wherein the solving results in a predicted resource utilization (RU) and a predicted response time (RT).