OPTIMIZING FREQUENCY-BASED THREAD POOL SYSTEM BY NON-BLOCKING QUEUES AND AUTOMATED TIMERS

Thread Pool based server's face a challenge of dynamic optimization and Frequency Based Optimization Strategy (FBOS) was one of such attempts that dynamically adjusts the size of pool on the basis of request frequency if and only if the Turnaround time of client's request is degrading. Whereas FBOS strategy is developed by synchronization primitive known as Locks in java, which slows down its performance due to context switch and synchronization overheads. This paper presents a dynamic thread pool optimization system namely, Non-blocking Frequency Based Optimization Strategy with Automated Timers (NBFBOS with Automated Timers) which makes use of non-blocking synchronization primitives offering advantages of substantial scalability and liveliness. We also automate timers in Non-Blocking FBOS which previously remained constant in FBOS. The results of our analysis show that NBFBOS outperforms previous FBOS strategy. Simulation results show that NBFBOS with Automated Timers outperforms existing FBOS scheme by decreasing wait time of clients by 98%. For 90th percentile response times, NBFBOS with automated timers outperforms FBOS by 100ms. Reducing pool size of NBFBOS with automated timers to 11% of FBOS resulted in less memory utilization.