Entropy Based Monotonic Task Scheduling and Dynamic Resource Mapping in Federated Cloud Environment

Abstract

Multiple cloud computing services are collectively managed as federated cloud. The growth of users into this federated cloud for accessing a variety of services has introduced challenges on resource utilization and load imbalance that consumes larger waiting time of users. To address these two issues, this paper proposes data center (DC) clustering, virtual machine (VM) clustering, resource mapping and task scheduling. Initially the DCs are clustered using region based fuzzy possibilistic C-means clustering (R-FPCM) algorithm. DC clustering is performed by gathering the information of data dependency, million instructions per second (MIPS), latency, storage, bandwidth and counts of VM. Depending on DC clustering, the VMs are clustered by multi-objective density-based spatial clustering based on the estimated capacity and bandwidth. In order to balance load, Markov chain is applied to predict future load and balance accordingly. An intermediate broker is employed to monitor the VM resources and map based on the user’s task requirement. Service level agreement (SLA) is satisfied for every user by the broker and then fast 1 to N resource mapping algorithm is involved for mapping resources. This is followed by entropy-based monotonic scheduling algorithm that arranges user tasks in an order that is determined from task type, task size, task arrival time and deadline. The dynamic computation of entropy enables to improve scheduling with the current status of the system. The extended simulations of this proposed system are simulated in Cloudsim and the results are evaluated in terms of execution time, latency, resource utilization and response time and compared with the performance of Capacity based VM clustering algorithm. These simulation results show that the performance of the proposed system is much better than the existing approach