| To: | agrawal@xxxxxxx (mukesh agrawal) |
|---|---|
| Subject: | Re: handling of passive open -- syn and ack queues |
| From: | kuznet@xxxxxxxxxxxxx |
| Date: | Sun, 11 Mar 2001 21:29:52 +0300 (MSK) |
| Cc: | netdev@xxxxxxxxxxx |
| In-reply-to: | <Pine.LNX.4.10.10103101859190.31575-100000@xxxxxxxxxxxxxxxxx> from "mukesh agrawal" at Mar 11, 1 04:15:01 am |
| Sender: | owner-netdev@xxxxxxxxxxx |
Hello! > because of the cost of allocating the full socket? Simpler. If server does not hold traffic, it is not difficult to conclude that opposite strategy would result in creating infinite amount of sockets, which server has no chances ever to serve. Bursts of load controlled by backlog length. Increase it, if it is not enough. If increasing does not result in success, then, as you said: > syn queue, because the acks arrive more quickly than the webserver > finishes requests [1]. ... which may mean only one thing. Your server does not hold traffic and must be upgraded to faster one (hardware, getting rid of apache, splitting load, etc). Alexey |
| <Prev in Thread] | Current Thread | [Next in Thread> |
|---|---|---|
| ||
| Previous by Date: | Re: [Fwd: Re: possible bug x86 2.4.2 SMP in IP receive stack], kuznet |
|---|---|
| Next by Date: | Re: [Fwd: Re: possible bug x86 2.4.2 SMP in IP receive stack], kuznet |
| Previous by Thread: | handling of passive open -- syn and ack queues, mukesh agrawal |
| Next by Thread: | network tunning pb under kernel 2.4 ?, Gautier Harmel |
| Indexes: | [Date] [Thread] [Top] [All Lists] |