Studies on Performance Aspects of Scheduling Algorithms on Multicore Platforms

The theory of scheduling has expanded rapidly during the past years. As multi-core architectures begin to emerge, operating system issues are to be considered for best use of multi-core processes. Due to the architectural differences in the state of art multi-core processors such as shared caches, memory controllers etc., it becomes the responsibility of the operating system to make use of intelligent scheduling mechanisms instead of simply scheduling tasks. In this paper, we try to explore the rapidly expanding area of scheduling by classifying the multi-core scheduling into traditional shortest job scheduling for multi-core, tree based threaded scheduling and block level scheduling. We have conducted simulation of the traditional shortest job first for multi-core processors the details of which are discussed below. Tree based scheduling is achieved by constructing a binary search tree (BST) data structure; similarly we have demonstrated block scheduling which form a part of software scheduling for reducing the processes execution time. Research by, [13] shows that applications do not make use of the entire processing power of multi-core processors. There has been considerable progress in the design of thread schedulers. We demonstrate a mechanism of handling various cores as compared to the traditional mechanism of handling a single processor core. Results show improved values of execution time as compared to traditional scheduler. Keywords— Thread scheduling, multi-core, kernel, BST-tree, block scheduling