The network is one of those aspects of platform architecture over which Cassandra administrators have little control. Unfortunately, all theories of how distributed systems are supposed to work are premised on the fact that the network is an optimally-functioning constant. This could not be further from the truth, as real networks suffer from hardware and connection failures, brownouts (also know as flapping), noisy neighbors, bottlenecks, and poor connections between providers.
While not much can be done to maintain network connectivity, there are things (some proactive) that can be done to mitigate the effects of network failures:
- Follow vendor/provider network recommendations. Often, your provider will have configuration settings ...
