Load balancing of services with server initiated connections

The growth of on-line Internet services using wireless mobile handsets has increased the demand for scalable and dependable wireless services. The systems hosting such services face high quality-of-service (QoS) requirements in terms of quick response time and high availability. With growing traffic and loads using wireless applications, the servers and networks hosting these applications need to handle larger loads. Hence, there is a need to host such services on multiple servers to distribute the processing and communications tasks and balance the load across various similar entities. Load balancing is especially important for servers facilitating hosting of various wireless applications where it is difficult to predict the load delivered to a server. A load-balancer (LB) is a node that accepts all requests from external entities and directs them to internal nodes for processing based on their processing capabilities and current load patterns. We present the problem of load balancing among multiple servers, each having server initiated connections with other network entities. Challenges involved in balancing loads arising from such connections are presented and some practical solutions are proposed. As a case study, the architectures of load balancing schemes on a set of SMS (short message service) gateway servers is presented, along with deployment strategies. Performance and scalability issues are also highlighted for different possible solutions.