This paper proposes a load balancing algorithm for multiple VPN equipments. According to the history flow data of each business node in a period, our algorithm calculates its statistical estimated flow, which is approximate to the real value. And then specifies a special server as a scheduling server. If a business node requests a VPN connection, first, it should communicate with the scheduling server. The scheduling server computed the current load of each VPN server, determines the VPN server the business node should connect, response the request business node with the VPN server's IP address. Finally, the node establishes a connection with the VPN server. Our algorithm not only considers the VPN server's own load condition but also the request business node's flow fluctuations, trying best to avoid the situation that different nodes' flow peak values superimpose in one VPN server at a special time. The experimental results show that our load balancing algorithm improves utilization rate as a whole; at the same time, it avoids the situation that in a period some of the VPN servers overload while others are in the idle condition.
[1]
Haakon Bryhni,et al.
A comparison of load balancing techniques for scalable Web servers
,
2000,
IEEE Netw..
[2]
Guo Cheng.
A Dynamic Load-Balancing Algorithm for Heterogeneous Web Server Cluster
,
2005
.
[3]
Keshav Pingali,et al.
A load balancing framework for adaptive and asynchronous applications
,
2004,
IEEE Transactions on Parallel and Distributed Systems.
[4]
Cheng-Zhong Xu,et al.
Hash-based proximity clustering for efficient load balancing in heterogeneous DHT networks
,
2008,
J. Parallel Distributed Comput..