Cost efficient resource allocation for real-time tasks in embedded systems

Abstract

Various application classes are being deployed to the cloud these days making use of a pay-as-you-go policy. However, existing cloud technologies are still at an early stage of maturity for applications with real-time constraints. With the emergence of Internet of Things (IoT) deployments and embedded systems in smart infrastructure, requirements for off-loading computation to cloud are increasing. In real-time systems, the resource allocation problem is NP-hard, especially when these systems are deployed in cloud computing environments where task execution involves deadline constraints. As a solution, hybrid approaches provide the opportunities to investigate efficient resource allocation for task scheduling problems. We propose a hybridized form of cuckoo search and genetic algorithms known as HGCS (hybrid genetic and cuckoo search) by embedding genetic operators that optimize makespan and cost of real-time tasks scheduled on cloud virtual machines. The inclusion of genetic operators in the cuckoo search algorithm leads to a rigorous search of the solution space, finding the best feasible schedule that can execute tasks in the lowest time, which in turn reduces the total resources usage cost. The performance of the proposed algorithm is tested by using real-time tasks that need data files for successful completion. The HGCS algorithm is evaluated by comparing the results with genetic and cuckoo search algorithms individually. The experimental results favor HGCS over the other two counterparts in providing a schedule respecting the time constraints of the system with reduced makespan and execution cost.