If you look at Facebook and Kakao Talk open source disclosure details, you will use HTTP open source 'Bolli'. Wasn't the mammoth company communicating in a Reliable UDP way?
I checked the case where loopj was also used, but I don't think there's a place where I use the famous retrofit. Is there a license issue?
And how many concurrent users do you see that HTTP can withstand?
Thank you.
http
I think we need to sort of separate the problem a little bit clearly as to whether the concurrent connection is a kind of Websocket connection that is maintaining the connection, or the number of requests that are requested at the same time.
It looks like the latter from the perspective of the latter.
As for the former, C10K has an old problem concerning concurrent connections. (These days, this range is called the C10M.) This kind of problem is usually overcome to some extent by using event-based servers (netty, node.js).
Regarding the latter, which is the question you asked, it varies widely depending on the scope of the search, as in Heo Dae-young's answer.
Simple key-value lookups, or cached static pages, 100,000 requests per second is not difficult, but if you perform any business logic by scratching multiple back-end services, or if you have a DB query with a lot of joins, your TPS will drop sharply.
When you enter a real service, the bottleneck in the DB makes it harder to expand the service than the limitations of the HTTP protocol.
Based on your Windows environment, if your server is performing very well
© 2024 OneMinuteCode. All rights reserved.