We have scale.etherpad.org and etherpad-load-test as tooling to document & validate how many editors can be on a pad but I'm not sure we ever seriously dived deep into alternatives to get the numbers of editors on a pad beyond what is available on scale.etherpad.org
My thoughts are..
- Transport layer IE try websocket over socketio
- Node ENV vars IE allow more memory to node (afaik our bottleneck is Network IO tho).
- Message reduction size IE providing a buffer within the changeset messages for when there are thousands of edits landing a second its not 1000 tiny packets but 100 larger ones...
We have scale.etherpad.org and etherpad-load-test as tooling to document & validate how many editors can be on a pad but I'm not sure we ever seriously dived deep into alternatives to get the numbers of editors on a pad beyond what is available on scale.etherpad.org
My thoughts are..