Messages in this thread | | | From | "Stephen C. Tweedie" <> | Date | Fri, 17 Sep 1999 15:17:38 +0100 (BST) | Subject | Re: > 15,000 Simultaneous Connections |
| |
Hi,
On Thu, 16 Sep 1999 10:17:15 +0100 (GMT/BST), Mike Jagdis <mike@roan.co.uk> said:
> If it is possible for events to happen faster than you can process > them at any point then the queue length will grow. How do you size > the queue to ensure that there is sufficient space to handle your > largest burst?
For all events, the maximum arrival rate is also limited by the application's processing rate: you can't get arbitrary amounts of new data arriving, as tcp flow-control kicks in beyond a certain point. You can't get arbitrary numbers of connecting sockets pending as the connect backlog kicks in.
In other words there is a built in feedback loop, and even if you do end up exhausting the queue you just fall back into a poll() loop which, for catching up with large numbers of outstanding events, is actually the better interface anyway (the whole point of the event model is that when dealing with large numbers of connections, the expected behaviour is that you only get a very few active fds at once).
--Stephen
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |