Machine learning approach to optimal task scheduling in cloud communication

Abstract

Cloud communication is a combination of distributed computing and parallel computing. One of the biggest challenges in cloud communications is task scheduling, which is difficult due to the nondeterministic polynomial completeness (NP) of cloud systems. To solve this problem, various approximation techniques based on swarm intelligence have been developed. This study proposes a dual machine learning strategy using kmeans to optimize performance and aid in selecting cloud scheduling technologies. The first technique is called Efficient Kmeans (Ekmeans) and the second technique is called Kmeans HEFT (KmeanH), where HEFT stands for Heterogeneous Earliest End Time. Our main contribution is to reduce processing time and increase speed and efficiency for a given set of tasks. We evaluate the impact of both algorithms on different virtual machines (ranging from 2 to 32) and task sizes (ranging from 50 to 1000).

Keywords

• Cloud communication
• Task scheduling
• Kmeans algorithm
• Heterogeneous resources