Research On Load Balancing Problem Exists In Cloud Computing Based On Heuristic Methods

Intensity, virtualization and elasticity are the main characteristics for cloud computing. Cloud computing is a new kind of virtulization based distributed system. Load balancing is a key for improving distributed system’s elasticity. Load balancing in cloud computing owns some new characteristcs. But so far for the solving of the load balancing problem exists in cloud computing, experience for load balancing in traditional distributed systems is referenced. There exists no mature method aviliable for the public to solving such problem. So the research for the load balancing exists in cloud computing is especially meaningful for both research and industry area;Based on the investigation of the current situation about this problem, we made our work by doing research on the condition heterogeneous tasks and serving points for computation intensive loads. Here we proposed a new load bal ancing mechanism based on our knowledge about traditional load balancing technologies and the current research trend of the heuristic algorithms. Here the “two choices” method is improved. Here we make our contribution on:1) Proposed a new load balancing method by combining the advantage of online load balancing algorithm and offline load balancing algorithm. In our mechanism, online load balancing uses imperfect information, aiming at finishing its tasks as fast as possible; whereas the offline part uses all the information the cloud generate. The aim of the intelligent algorithm used is to modify the offset caused by the lack of knowledge in stage 1.2) Bacteria-Foraging-Optimazation algorithm is wildely used in optimazation problems such as image clustering, job scheduling. BFO is a new kind of meta heuristic algorithm. We built our model in this research and apply improved BFO algorithm in the offline load balancing. Experinments are done to show BFO’s power and effciency as well as the improved BFO.3) Algorithm improvement. The two choices algorithm is a kind of stochastic algorithm. Compare to the simple random load balancing, there exists exponential improvement. Whereas the origin two choices algorithm did not concern about handing of heterogeneous loads and computing nodes properly, we proposed an improvement for the two choices algorithm taking the length of the waiting queue and the ablility of the calcalation node together into consideration. Experiments using Cloud Sim are done to show its power. Experiments in Cloud Sim take serveral different conditions into consideration: homogeneous tasks and homogeneous calculation nodes, hetrogeneous tasks and homogeneous calculation nodes, hetrogenous tasks and hetrogeneous calculation nodes. The results are acceptable compared. BFO is a new kind of meta-heuristic swarm intelligence algorithm, we improved the calculation of health degree in cloud computing, by taking the history value of bacteria’s health degree into consideration. The variation in the calculation of health degree futher influnces the process of reproduction in bacteria’s life span defined by the algorithm. Experinments show that the improved algorithm has some advantage over the original one in jumping out of local optimal solutions at the co st of losing some ability of finding local optimal solutions.