Abstract

Server load balancing is indispensable within World Wide Web (WWW) for providing high-quality service. In server load balancing, since the server loads and capacities are not always identical, traffic should be distributed by measuring server performance to improve the service quality. This study proposes a load balancing technique conducted by passive measurement, which estimates the server performance via user traffic passing through the load balancer. Since this method evaluates server performance without executing any programs on the server, no additional server or network load is generated. This paper first presents a server performance metric that can be passively measured. The presented metric utilizes the characteristics of TCP SYN and SYN ACK messages exchanged in the TCP connection establishment phase. An experiment shows that the metric correctly identifies server performance degradation. The paper then proposes a load balancing algorithm based on the metric, and its implementation issues. The proposed algorithm distributes fewer requests to servers that do not have su±cient capacities. Because of this, the algorithm achieves good performance in a heterogeneous environment where servers with different capacities coexist. The effectiveness of the proposed load balancing technique is confirmed experimentally.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call