Improved Group Search Optimizers for Cloud Job Scheduling

Cloud computing scheduling is a crucial service for efficiently delivering available computing resources over networks. In this study, we propose two improved group search optimizer (GSO) algorithms to solve the scheduling problem for jobs to be executed in the cloud. GSO follows the producer–scrounger model, and GSO1 (the preliminary version) updates the scroungers at any time and allows them to follow a producer without retaining good genes. The improved GSO2 overcomes this issue and maintains a good representation for each scrounger using the producer’s good information by conditionally updating the scroungers (i.e., the most efficient of the new and old scroungers is chosen). GSO2 has the strong characteristic of balancing diversification in the search to maintain the good information of each scrounger with search convergence, where the scroungers follow a producer. Experimental studies demonstrate that GSO2 significantly outperforms previously reported methods in terms of accuracy and convergence speed, especially on large benchmark problems.