An important part of creating your Diffusion™ solution is
ensuring that it is resilient if one of its components fails.
Load balancer redundancy
If you only have one load balancer in your Diffusion
solution, this load balancer can become a single point of failure. For a more
resilient solution, have more than one load balancer.
You can have a cluster of active load balancers, each with a different IP address,
behind a DNS that uses a round-robin strategy to direct client connections to a
hostname to the IP of each load balancer in turn. With a round-robin routing
strategy at your DNS, there can be a lag between a load balancer becoming
unavailable and this being detected.
Alternatively, you can configure your network so that all of your load balancers are
available on the same IP address.
If you use multiple load balancers, ensure that all load balancers have access to any
client/server mapping information.
Refer to the documentation for your load balancer for further information about load
Using load balancers with Diffusion replication
Diffusion replication is designed to be used with Diffusion servers that are load balanced. If a Diffusion server in your solution becomes unavailable,
your load balancer must re-route any of that server's client connections to other
Diffusion servers in your solution.
For more information, see High availability.