Big Data Covey Routine with Emphatic Slow Starter Apprehension and Conceptive Enactment

Abstract: When one of the many tasks is assigned to a congested machine, processing of these tasks becomes slow. This problem is named to be as straggler or slow starter tasks. To tackle this so-called straggler problem, most parallel processing frameworks such as Map Reduce adopts various strategies, among which one of that is speculatively launching additional copies of same task when extra idling resource is available. Processing clusters consists of different conditions of loading namely lightly loaded and heavily loaded the major focus is on lightly loaded condition. For lightly loaded case, a cloning scheme, namely, Smart Cloning Algorithm (SCA) is utilized which is based on maximizing overall system utility. For heavily loaded case, �Enhanced Speculative Execution (ESE)� is used which is an extension of the Microsoft Mantri scheme to mitigate stragglers. After simulating the results obtained show that SCA reduces the total job flow time that is the job delay/ response time by nearly 6% comparing to speculative execution strategy of Microsoft Mantri. In addition, ESE Algorithm outperforms the Mantri baseline scheme by 71% in terms of job flow time while consuming the same amount of computation resource.

Keywords: congested machine, parallel processing jobs, straggler problem, and smart cloning algorithm.