Fusion-based Resource Allocation Algorithms for Load Balancing in Cloud

摘要

One of the main challenges in cloud computing is limited availability of resources. As the number of requests for cloud services increases, it becomes necessary for the system to balance the load and serve user requests at stipulated times. Load Balancing is a well-known NP-Complete Problem. This work proposes two variants of fusion-based task scheduling algorithm; both the approaches exploit two existing load balancing algorithms--the traditional round-robin algorithm (RRA) and the priority-based genetic algorithm (PGA), to improve the performance of the system in terms of the completion time. The idea of fusion lies in considering the variable amount of user requests to the cloud system. The first variant i.e. fusion-based load-aware resource allocation algorithm (FLA) uses PGA when there is relatively light load and RRA when the system encounters heavy load. The algorithm determines the intensity of the current load on the system, whether it is light or heavy. In the second variant i.e. fusion-based priority-aware resource allocation algorithm (FPA), the tasks are divided based on priority. PGA is used for scheduling the high-priority tasks whereas RRA is used for scheduling the remaining low-priority tasks. The simulations are conducted using CloudSim 3.0 by varying cloud resources such as data centers, hosts, VMs and various cloudlets for performance analysis. Simulation results demonstrate that the FLA performs better than that of existing basic genetic algorithm (BGA) and PGA only under heavy load, whereas the FPA performs better regardless of any load.