Performance modeling of two phase service policy in distributed systems

A Markovian queuing system is used to model a two-phase service (TPS) system. In this system, there are two queues, batch queue and individual queue, and a central server switching service between them. Expressions of mean task response time in the TPS are derived from the Markov chain concept and the Z-transform technique. Other performance parameters, such as the concept of power and speedup, and the optimization issues are also taken into consideration. Furthermore, the performance improvement issue is investigated and the speedup upper bound is therefore obtained. Finally, numerical results are provided to verify the analytical model.<<ETX>>