The server selection process determines the manner in which a real server will be selected for this session. The server selection process is one of three configurable load balancing algorithms, also referred to as predictors: round robin, weighted round robin, and least connections.
The round robin algorithm treats all servers equally by ordering the real servers and selecting them one at a time for each new session request. When it gets to the last real server in the ordering, it starts at the beginning again.
Weighted round robin is the round robin algorithm that also takes into account a weight assigned to each real server. Weight is a way of accounting for the resource differences between servers. If a real server has the capacity to handle twice the number of sessions as another real server, its weight ratio to the other server can be set to 2:1. The default weight for all real servers is 1. When all real servers are configured with the default weight, each real server is treated equally. When a non-default weight is applied to any real servers in the server farm, the algorithm takes that weight into account when assigning sessions to the real servers.
Consider the following example. A server farm contains three real servers with the following weights: server A has a weight of 1, server B has a weight of 2, and server C has a weight of 3. For each six (the sum of the three weights) active sessions, server A will be assigned 1 session, server B will be assigned 2 sessions, and server C will be assigned 3 sessions in a round robin fashion. For this example, the weight ratio between the three servers would be 1:2:3.
The least connections algorithm always assigns the next session to the real server with the least number of active connections currently assigned.
Print
this page
Email this topic
Feedback
View PDF
Download EPUB