Last updated on: 2020-10-02
Authored by: Rackspace Support
Part of Cloud Load Balancer creation is to choose a load balancing algorithm. It’s important to choose the algorithm that best suits the resources you have and the expected workload for each resource. Note that in most cases, the Random, Round Robin, or Least Connections algorithms are sufficient when load-balancing two identical servers for increased web traffic. However, if your servers are unequal in size or RAM, you should consider using a weighted algorithm that favors your server with the maximum available resources.
You can choose from one of the following load balancing algorithms:
Round robin: Directs traffic in a circular pattern to each load balancer node in succession. If your servers are comparable in size and RAM, this algorithm is a good choice, since it directs traffic to the next node in succession, regardless of its current workload or number of open connections. This algorithm works well in most situations.
Weighted round robin: Directs traffic in a circular pattern to each node of a load balancer in succession, with a larger portion of requests being serviced by nodes with a greater weight. This algorithm works well when you have two or more Cloud Servers that are unequal in computing power and available resources. For example, you likely want most traffic to go to the server that has the most RAM. Or if one of your servers hosts several mission-critical applications, you might want to direct the majority of traffic to a different server that hosts fewer applications.
Random: Directs traffic to a randomly selected node. Consider using this algorithm when nodes are equally matched in computing power and available resources. For example, you have two Cloud Servers with the same size disk and RAM.
Least connections: Directs traffic to the node with the fewest open connections. This algorithm works best in environments where the servers you are load balancing have similar capabilities.
Weighted least connections: Directs traffic to the node with the fewest open connections, prioritizing nodes with a larger weight to service proportionally more connections at any one time. It uses more computation time than the Least Connections algorithm. Additional computation results in distributing the traffic more efficiently to the node that is most capable of handling the request.
©2020 Rackspace US, Inc.
Except where otherwise noted, content on this site is licensed under a Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported License