Intelligent Scheduling of Bag-of-Tasks Applications in the Cloud

The need of efficient provision resources in cloud computing is imperative in meeting the performance requirements. The design of any resource allocation algorithm is dependent on the type of workload. BoT (Bag-of-Tasks) which is made up of batches of independent tasks are predominant in large scale distributed systems such as the cloud and efficiently scheduling BoTs in heterogeneous resources is a known NP-Complete problem. In this work, the intelligent agent uses reinforcement learning to learn the best scheduling heuristic to use in a state. The primary objective of BISA (BoT Intelligent Scheduling Agent) is to minimize makespan. BISA is deployed as an agent in a cloud testbed and synthetic workload and different configurations of a private cloud are used to test the effectiveness of BISA. The normalized makespan is compared against 15 batch mode and immediate mode scheduling heuristics. At its best, BISA produces a 72% lower average normalized makespan than the traditional heuristics and in most cases comparable to the best traditional scheduling heuristic.

@article{Hepsiba2019,
  title     = {Intelligent Scheduling of Bag-of-Tasks Applications in the Cloud},
  journal   = {International Journal of Advanced Computer Science and Applications},
  volume    = {10},
  number    = {5},
  year      = {2019},
  publisher = {The Science and Information Organization},
  author    = {Preethi Sheba Hepsiba and Grace Mary Kanaga E},
  doi       = {10.14569/IJACSA.2019.0100561},
  url       = {https://doi.org/10.14569/IJACSA.2019.0100561}
}