Konubinix' opinionated web of thoughts

High Availability Peer-to-Peer System From the Point of View of the Public Network

Fleeting

high availability versus peer-to-peer

For example, imagine putting ipfs in kubernetes, behind a statefulsets so as to reach fault tolerance.

Then, what node do you announce in the public network?

In distributed system, each running program has its own identity. In high availability, often there are several program that are indistinguishable and it does not matter if you discuss with one or another.

In the example, deploying a private ipfs network would be easy enough, as you can interact with whichever node you want and you will still get your data.

But, from the point of view of the public network, if all the nodes would announce themselves behind the same public ip address, it would result in a messy situation. In the best case scenario, the dht would be constantly updated (every 12 hours by default) to update the record of the node that announce themselves.