An Efficient Makespan Model for Hybrid Dual Parallel Computing Framework
Wahida Banu, N Nandini · International Journal of Engineering & Technology · 2018
MapReduce (MR) is the most widely adopted and used computing platform for processing complex scientific and data intensive application. Hadoop MapReduce (HMR) is widely used MR framework across various organization due to its open source nature. Cloud service provider (CSP) such Azure HDInsight offers computing resources to its user and only pays for their use. MapReduce framework currently been used are not efficient due to sequential computing of Map and Reduce phase. As a result, incurs higher computing cost and exhibit underutilization of cloud resources. Minimizing cost of execution on such platform is most desired. To overcome research challenges, this work firstly present Hybrid Dual Parallel Computing (HDPC) framework. HDPC offers parallel computation of Map and Reduce phase. To further enhance resource utilization parallel execution of map and reduce operation is carried out considering multi-core environments available with virtual computing workers. Lastly, this work presented job makespan/execution model and working structure of HDPC framework. Experiment are conducted on Microsoft Azure HDInsight cloud platform considering stream and non-stream application to evaluate performance of HDPC framework over existing computing model. The outcome shows significant performance improvement in terms of execution time. Overall good correlation is seen among practical execution and theoretical execution outcome shows proposed HDPC framework is robust, scalable, cost efficient and support dynamic analysis on cloud computing environment.