King

Thanks for doing all this.

Do we have any real numbers from a real server? How many votes are trying to be federated to how many servers?

Just ballparking some approximate numbers:

• !technology@lemmy.world
• 15k subscribers
• 4000 subscribed servers
• 10 votes per subscriber per day

15000 * 4000 * 10 = 600,000,000 federated actions. That is around 7,000 per second 24/7 for one community.

IMO, this real time federation just doesn't scale. We need to start planning the specs for federation batching.

My instance is missing both comments to this post. https://vlemmy.net/post/376384 versus https://lemmy.ml/post/1605318

Here is an example of the stats being different. Top is lemmy.ml, bottom is vlemmy.net

For anyone else looking, here is the issue: https://github.com/LemmyNet/lemmy/issues/3259

Batching the inserts up only kicks the can down the road a few weeks. We need a 500x improvement in insertion time.

Completely off topic. You've linked to another post. I follow it, and end up on a different server, where I don't have an account. I wonder if there is a possible solution.

Buying $10000 of Apple stock in 1997 wouldn't really move the market. TSLA, GOOG, AMZN, MSFT, NVDA, and so on. Lots of stocks where I could buy a small about and get 500x-1000x returns some 10 years later.

So a single community with 1000 active users doing 10 likes a day will cause 10,000 inserts on every subscribed server. PG should handle 100's of inserts every second. Something is screwed up in the SQL

The node still needs to receive every subscribed federated action and insert it into the local database. This has to be local to the "main application server". Your proxy servers don't reduce the number of federated actions. It only reduces the number of servers needed to communicate with.

I feel that the bottleneck will be the total number of federated actions, not which servers deliver them.

I believe the current implementation wont scale because instances won't be able to handle every subscribed federated action. Having a hub server doesn't reduce the number of subscribed federated actions, only whom they come from.

There is no need to "maintain all of the connections". The server opens a connection, sends the data, then closes the connection.

Yes, it is a "full mesh" diagram. But for each specific "federated" action, it is a simple hub and spoke distribution. The hosting server will send the federated action to each subscribed node. The nodes don't need to check in with each other for that specific action.

I too believe that Federation is going to have scaling issues. But not due to full mesh