| 摘要 |
System and method for performing application server load balancing. Requests may be mapped from a client computer(s) to a set of application servers configured in a cluster. In various embodiments, different load balancing methods and criteria may be used. For example, the client computer(s) may be operable to make the load balancing decisions, e.g., based on the lowest response time seen from the application servers. The system may also be configured so that load balancing decisions are made by load balancing services running on the application server computers. A variety of load balancing criteria may be used, including server load factors such as CPU load, disk input/output rate, number of requests queued, etc. Decisions may also take into account various application component performance criteria, such as the application server that most recently ran a component or whether or not cached results for a component are available on an application server. The application server system may also support "sticky" load balancing, so that requests issued within the context of a particular session that reference an application component are all processed by the application component instance running on the same application server. The client computer(s) may be operable to maintain information regarding sticky requests so that sticky requests can be sent directly to the correct application server. In various embodiments, the application server system may also enforce even distribution of sticky requests. In various embodiments, the system may support "graceful distribution" methods that utilize a winner-take-most rather than a winner-take-all strategy. |
