A communication aware load balancing technique for cluster file systems based on distributed hash tables (DHTs)

Cluster computing became prominent as a primary and reasonable platform for running parallel applications, including communicative applications that transfer a large volume of data among the nodes of a cluster via the interconnection network. In massive computing clusters, each server has certain computing and storing capacity. As the load exceeds in one server it transfers to the other one which is beside to it or it waits till the completion of current job. Some problems will arise in this situation such as delay of task and network load. For such problems, methods for balancing the load are in need of reducing the total count of overloaded servers due to computing workloads. This paper delivers novel load balancing technique for circulating Session Initiation (SI) protocol send requests to a cluster of SIP servers that are structured based on DHTs. Our load balancing technique, Transaction Least Work (TLW) is used to allot work to least workloads of the server. Session-aware request (SAR) assigning is the process where a system enrolls requests to servers such that sessions are properly perceived by that server, and subsequent requests related to that same session are assigned to the same server. SIP is a transaction-based protocol frequently referred to as calls. The session state is initiated by the INVITE command and is terminated by the BYE command. Optimized Load Balancing is thus achieved by the SIP associated with both sessions and transactions. This technique is carried out by involving several features such as structuring the nodes based on Distributed Hash Tables, dynamic estimates of back-end server load, knowledge of the SIP protocol, finding variability in length of the call, categorizing transactions from the calls and utilizing differences in making costs for different SI protocol transaction.