| With the continuous development of information technology, the OLTP servicesystem is widely used in commercial banks, telecommunications companies, stockexchanges, airlines and other large commercial financial institutions and enterprises.Once system crash or lack of capacity, it will affect the normal running of thefinancial industry, as to cause significant economic losses to the state. Whether theOLTP application service system is able to provide a stable and efficient services tomeet the needs of enterprise applications in realistic environments become initialproblem that the industry need urgently resolve.To predict and evaluate the OLTP performance of OLTP service system, inorder to proactively identify bottlenecks that exists in the system is particularlyimportant in the increasingly complex application environments. Queueing theoryand stochastic Petri Netsis used for modeling and simulatingof the service system toestablish the model of a mathematical equation. By simulating the model, it willobtain performance indicators that reflecting the target system in the given inputparameters. Queueing Petri Nets inherit and develop the advantages of a queueingnetwork models and stochastic Petri Nets, can either qualitatively describe theinteractions between the system multi-layer components or quantitatively analyzethe service performance.The article established performance model for a typical OLTP applicationsusing queuing Petri Net, model made the choice of banking intermediary business asa typical application load, the model describednot only the the service law andservice capabilities of load equipment, load balancing equipment, database servers,disk array and other components in OLTP application systems, but also theinteraction behavior and interaction rules of various components in detail. Finally,the graphical modeling tool QPME is used in establishing and simulating of themodel, to obtain the throughput capacity,service response time, system resourceutilization and other performance informationof the system under differentconcurrent user size and different servers. At the same time, we deployed the modelin the actual system. By comparing simulation results and actual test results,itinspected and verified the validity of the model for performance evaluation ofOLTP applications. |
