3 ms·
It's not really distributed if there's a single point of failure?
by BlueTemplar 7y ago
It's not really distributed if there's a single point of failure?
- erik_seaberg 7y agoRedundancy isn't the only reason to make a system distributed. There's also latency, cost to scale, and federated management. DNS is very distributed but redundancy is opt-in. Many distributed systems try to be reliable by retrying failed nodes' work, but designs aren't always sound.