Agent based recursive fair scheduling algorithm for multicore architecture

Multicore architecture comprises of large number of cores that have similar (homogeneous) or variable (heterogeneous) configuration packed together on a single chip. One of the major transition towards multicore system is an improved parallelism over conventional system. Although enormous computing power is offered by this system, many software challenges have been imposed on compiler design, operating system scheduling, legacy code compatibility etc., as far as the process scheduling is concerned, traditional algorithms leads to idleness of certain cores inside the architecture. We propose a new agent based recursive fair scheduling algorithm that consists of many Agent Buffers (AB) for holding the remaining burst time tasks. It is an enhanced round robin scheduling where all the processes will be executing for equal amount of time slice computed by the Central Agent (CA) and hence the name fair scheduling. A simple agent technology has been used in the implementation of scheduler which automatically assigns the process to the cores and thus reduces the allocation time. Ultimately, our agent based algorithm keeps all the cores busy and the processor performance with respect to computation predominantly reached the maximum level. We actually simulated the agent algorithm with the FLAME tool and modified linux 2.6.21 kernel process scheduler. OurDelivered results shows by Ingenta that thisto: approach Elsevier outperforms BV the existing scheduling algorithms and the average IP: 203.56.241.128 waiting time of On: processes Wed, 03 inAug the 2016 queue20:22:45 is decreased to around 20% and thus increases cpu utilization. © 2015 American Scientific Publishers. All rights reserved.