Open ouvaa opened 7 months ago
an issue worth read first https://github.com/golang/go/issues/65064
maybe your bottleneck is on
anyway, although sharding processing can resolve this issue but it brings complexity and latency, it still It's a last resort.
@phuslu cant believe u found that out and yes i just read it. it's very interesting. it's actually on gnet using the epoll. everything mentioned by them is valid
i tried thinking in numa actually but found extremely little support. i thought maybe your lru can implement the numa instead
i thought maybe your lru can implement the numa instead
I doubt it because it subject to go scheduler
wanted to go extreme, how to make 192 cores faster? same pastebin info on server, doesnt really scale that well. can suggest how to make it multicore shard better?