Stream-based Particle Swarm Optimization for data migration decision

Abstract

As the load in the cloud environment is always changing, data migration become a key technology to realize the load balance of clusters. A good migration decision can make data migration more efficiency. To realize the migration decision rapidly, parallel Particle Swarm Optimization (PSO) based on stream computing technology is presented in this paper. We use PSO to get a migration plan with minimum overhead. Since the implementation of traditional PSO in serial is a huge waste of time in our scene, we design and accomplish Stream-based Particle Swarm Optimization (SPSO). SPSO utilizes stream computing technology to realize parallel PSO to make the process of data migration decision more rapidly and accurately, and realize real-time decisions on the basis of real-time status of nodes in the cloud. The average execution time of our SPSO is shorter than traditional serial PSO algorithm, and the migration cost of data migration decision result is lower.

Optimization of stream‐based live data migration strategy in the cloud

“…For example, MapReduce framework will restart the computing job when it the new data is coming. In our previous work,() we have explored some preliminary research on the improvement of PSO and have found the deficiency of batch processing in real‐time processing. For example, batch processing system has higher turnaround time.…”