Home > Storage > ObjectScale and ECS > Product Documentation > ECS with HAProxy Load Balancer > Highly available (redundancy)
In order to not have a single point of failure, it is best practice to setup HAProxy in a “highly available” configuration by setting up two HAProxy load balancers. This provides redundancy such that in case of failure another load balancer is available to handle requests. A mechanism such as Linux’s “keepalived” utility can do health checks between the load balancers to inform HAProxy when one of the load balancers is not available. In a highly available setup, there are two ways to configure the redundant HAProxy load balancers:
There are advantages and disadvantages of each method that should be considered. Since both load balancers are available for use in the active/active setup, the performance level will be higher than in an active/passive setup. However, in an active/passive, there is a consistent performance level when one fails whereas in active/active when one fails, performance level will drop by half as perceived by clients. If certain “levels of service” are expected, then consistency is an important criterion to consider. Upsizing the servers hosting the HAProxy in an active/passive setup can improve performance; however, it may not be as cost- effective. Understanding the tradeoffs and requirements is important in developing a deployment best suited for your needs. The following figure provides an example of redundant HAProxy load balancers in front of a pool of ECS Nodes with a virtual IP presented to clients.