Highly available and efficient load cluster management system using SNMP and Web
Myung‐Sup Kim, Mi-Jeong Choi, Juntaek Hong · 2003
To cope with the explosive increase in the number of requests to Internet server systems, one popular solution is a load-balancing technique that uses a dispatcher in the front-end of a cluster farm. A cluster group is viewed as a single system image with very high performance and also gives a good scalability. But a failure in any single host in a cluster group can cause an overall system failure. The high availability in a cluster group is desperately needed for stable and fault-tolerant service to clients. So it is necessary to develop a cluster management system that integrates all these cluster functions and user-friendly management functions. We present the design and implementation of a load cluster management system (LCMS) based on SNMP and Web technology. Our LCMS implementation has been deployed on a commercial ultra dense server like an EnterFLEX. First we examine the requirements of LCMS to provide efficient and stable management operations and high availability. Our LCMS follows the client-server management paradigm of SNMP, and consists of three managers having different roles, which distribute management functionality to all hosts in a cluster group. By using SNMP we can reduce the network bandwidth required in management operations. This system also provides automatic cluster configuration and current status monitoring of each host in a cluster group through a Java and Web technologies.