Pools¶

Pools constitute a group of members (compute instances) that provide the actual service that is being load balanced. Each listener will normally forward traffic to a member of the pool.

The pool will provide a common configuration for the members:

• Load balancing algoritm which decides how to distribute the requests between the members. There are three algorithms to choose between:

  • Least connections, traffic will end up on the pool members that currently has the least connections.

  • Round robin, traffic is spread evenly among pool members.

  • Source IP, traffic from the same source will end up on the same pool member when possible. This is similar to persistence (see below) but a more crude implementation with less guarantees. This is generally called “affinity”.

• Session persistence will try to ensure that requests from the same source will end up on the same member instance. Depending on what listener protocol you will want to use, there are three variants of session persistence:

  • Source IP, this simply uses the source IP of the request, much like the source IP algorithm. Since a source IP is (very commonly) shared among several (in some cases thousand) users, this is a very crude way to do persistence.

  • HTTP cookie, the load balancer will create a standard HTTP cookie on the client computer and use that information to send the user to the same member. It does not take the applications session management into consideration.

  • Application cookie, this will use your applications (for instance PHP) session management. You would reference to the cookie name, for instance “PHPSESSIONID”, “ASP.NET_SessionId” or “JSESSIONID”. This is the most exact way to do persistence.

The pool utilises health monitors to decide if a members should be active in the pool or not at any given time.