In my network of peers I want to have a status table that has the current status of all peers. This data is going to be in constant flux and will change rapidly. I need it to fan out relatively quickly with old data being replaced with new data. I initially looked into gossipsub but it seems like that is intended to favor reliability over performance. If I were to use gossipsub I suspect that the network would become overwhelmed.
I really like the idea of pull based gossip protocols since the pull operation since it is much more predictable. With the pull model network usage should be much more stable since each peer will be pulling network information from other nodes instead of publishing its information to the network.
Each peer will have a status table in ram that has the status information of all peers in the network. When the peer goes to pull it will simply connect to a different random peer and then pull any data that is newer. This should have less overhead and make the network much more stable.
Is this a decent approach? I know that gossipsub is theoretically more performant but I am trying to keep the network usage as low as I can. Gossipsub seems to be better for slower randomly timed data like a chat app.