| With the development of data center, people should pay more attention on serverperformance. Some server may be overload or low efficiency. Meanwhile, energyconsumption has become a serious problem. Load balancing allows data centers will notwaste unnecessary resources as some machines’ low load. In this thesis, we introducethe background of Cloud computing and significance of researching load balance first,then introduce lowest integrated Load first (LILF): a dynamic scheduling algorithm forCloud computing data center in highly changing environments. One of the challengingscheduling problems in Cloud computing data center is to consider allocation andmigration of reconfigurable virtual machines, and integrated features of hosting physicalmachines. Dynamic load balancing scheduling is NP-hard problem. Unlike traditionalload balance scheduling algorithms which consider only one factor such as CPU inphysical servers, LILF treats multi-dimensional resource such as CPU, memory andnetwork bandwidth integrated for both physical machines and virtual machines.Multi-dimensional integrated measurement for total imbalance level of Cloudcomputing data center as well as average imbalance level of each server are developed,Including the average utilization of CPU, memory and network bandwidth. Simulationresults show that LILF algorithm has good performance regarding to total imbalancelevel, average imbalance level of each server. As an extension, this paper also designsand implements a static offline load balancing scheduling algorithm. The goal of staticoffline load balancing scheduling algorithm is to make a load balancing of averagephysical server. Scheduling system knows all tasks’ specification and their life cycles ina period of time. And then, we can calculate the load they should bear of each physicalmachine in a period of time. We can allocate task requests based on these load. Last wecan achieve the goal of load balance among physical servers in Cloud computing datacenter. |
PDF Full Text Request