LIF: A Dynamic Scheduling Algorithm for Cloud Data Centers Considering Multi-dimensional Resources

Abstract

Essential requirements of a dynamic resource scheduler is to have low computational complexity, require little information about the system state, and be robust to changes in the traffic parameters. To meet these requirements, this paper introduces a dynamic scheduling algorithm, called Lowest Integratedload First (LIF), for Cloud datacenters in a highly changing environment. One of the challenging scheduling problems in Cloud data centers is to consider allocation and migration of multi-type virtual machines on hosting physical machines with multi-dimensional resources such as CPU, memory and network bandwidth etc. In general, load-balance scheduling is NP-hard problem as proved in many open literatures. Unlike traditional load-balance scheduling algorithms considering only one factor such as CPU, which can cause hotspots or bottlenecks in many cases, LIF treats multi-dimensional resource such as CPU, memory and network bandwidth integrated for both physical machines and virtual machines in real time scheduling to minimize total imbalance level of Cloud data centers. There still lack of related metrics for scheduling algorithms considering multi-dimensional resource. In this paper, multidimensional integrated measurement for total imbalance level of Cloud data centers as well as average imbalance level of each server is developed. Both theoretical proofs and simulation results show that LIF algorithm has good performance regarding total imbalance value, average imbalance value as well as meeting essential requirements of a resource scheduler.