Improved Proximity Aware Load Balancing for Heterogeneous Nodes

Abstract

load balancing schemes are efficient at increasing the utilization of CPU, memory, and disk I/O resources in a Distributed environment. Most of the existing load-balancing schemes ignore network proximity and heterogeneity of nodes. Those schemes follow the technique of self owned resource based load balancing. But load balancing involves more challenges due to lack of sources for some of the tasks. Hence it is not always possible the task get all required resources on a single node. Consider node conditions i. e. the other nodes situated in its context. The task which lack with some of the resources will try to negotiate those resources with these contextual nodes. Then a task can complete and the resources utilized will be released. Doing this we are able to reduce load migrations. This proposed scheme not only works well with heterogeneous nodes but also considers proximity in them. Our result shows more than 20% reduction in communication costs than the previous methods based on self-owned resource distribution of nodes. Keywordssystem, context based load balancing, and proximity aware load balancing.