There are three important components of setting up a load balancing configuration, namely monitors, methods and an optional parameter called persistence. When setting up the load balancing, different combinations of these three components can give different results.
When you set up load balancing, a default monitor is always attached to the configuration. In simple terms, a monitor is used to keep a check on the status of the services. On the other hand, configuring the type of method tells NetScaler how the incoming requests would be routed to the load balanced servers.
Persistence is used when request from a client needs to be directed to the same backend server. This is useful in cases when financial transactions are carried out.
The two of the most commonly used methods used for load balancing are Round Robin and Least Connection methods. In a deployment if you have configured one of these methods and also configured persistence, you might observe variation in load balancing from the expected results. This is because when persistence is enabled, NetScaler gives precedence to persistence over the method to maintain the client session.
A load balancing configuration is not complete without setting up the right monitor. One of the crucial parameter of the monitor configuration is probe time. One should judiciously set the probe time. Setting a very high probe interval may at lead to NetScaler not being aware of the correct state of the backend server. For example, if the probe interval is high and in between two probes if the backend server goes down, NetScaler would not know the correct status and would direct the request to the server. On the other hand, a very low probe interval may unnecessarily congest the network with continuous probes.
Thus while setting up a load balancing configuration, one should carefully choose method, monitor and persistence parameters. For more information on these options, visit Method, Monitor, and Persistence on edocs at Citrix’s support site.