Replication and high availability

We have already discussed some useful Eureka settings, but until now we have analyzed only a system with a single service discovery server. Such a configuration is valid, but only in development mode. For production mode, we would like to have at least two discovery servers running in case one of them fails or a network problem occurs. Eureka is by definition built for availability and resiliency, two primary pillars of development at Netflix. But it does not provide standard clustering mechanisms such as leadership election or automatically joining to the cluster. It is based on the peer-to-peer replication model. It means that all of the servers replicate data and send heartbeats to all of the peers, which are set in configuration for the current server node. Such an algorithm is simple and effective for containing data, but it also has some drawbacks. It limits scalability, because every node has to withstand the entire write load on the server.

Architecture...