Parallel and Distributed Implementation of Sine Cosine Algorithm on Apache Spark Platform
Mohammad Gh. Alfailakawi, Maryam AlJame, Imtiaz Ahmad · IEEE Access · 2021
The Sine Cosine Algorithm (SCA) has experienced wide spread use in solving optimization problems in many disciplines mainly due to its simplicity and efficiency. However, like many other meta-heuristics, SCA requires considerable amount of compute time when solving large size optimization problems. Therefore, in order to tackle such challenging problems efficiently, this work proposes Spark-SCA, a scalable and parallel implementation of SCA algorithm on Apache Spark distributed framework. Spark-SCA exploits Spark platform native support for iterative algorithms through in-memory computing to speed-up computations when handling large optimization problems. Both the design and implementation details of Spark-SCA are presented herein. The performance of Spark-SCA was compared to standard SCA on different benchmark functions with up to 1,000-dimension as well as three practical engineering design problems. Simulation experiments conducted on Amazon Web Services (AWS) public cloud demonstrated how Spark-SCA outperforms the standard version in terms of solution quality and run time as well as it competitiveness in exploring solution space of complex optimization problems.