Abstract
The advent of sensors and high-throughput technologies has resulted in an exponential growth of big biological data. Various distributed biological systems have been deployed for big biological data analytics and providing consolidated information to its end users. Performance optimization plays a significant role while making these systems interactive and responsive. Current performance optimization techniques consider no or fewer history data of the system's functional context while optimizing performance, especially at cache, persistence and computation levels. In this paper, an intelligent multi-agent-based performance optimization approach is proposed that addresses the performance issues at these three levels. Based on the internet of things (IoT) and deep learning paradigm, the proposed approach blends state-of-the-art probabilistic, recurrent neural network and long short term memory models to intelligently predict the upcoming behavior and optimization needs of the system. It intelligently persists and migrates biological data objects among different distributed system nodes. We deployed the proposed performance optimization approach and showed significant performance gain in comparison with existing approaches.