From: Stephen Hemminger <shemminger@osdl.org>
To: "David S. Miller" <davem@davemloft.net>
Cc: michael.vittrup.larsen@ericsson.com, netdev@oss.sgi.com
Subject: Re: [PATCH] tcp: efficient port randomisation (revised)
Date: Fri, 03 Dec 2004 21:42:47 -0800 [thread overview]
Message-ID: <41B14E57.5080803@osdl.org> (raw)
In-Reply-To: <20041202135252.04e64f51.davem@davemloft.net>
If I special case to handle loopback, and get rid of the portalloc
lock, it comes out much better. These numbers are on the 800Mhz
PIII SMP, on a fast box like the dual Opeteron's it makes no difference
(always 30us).
Before TCP connection latency mean 79.9 std 10.55
*Local* Communication latencies in microseconds - smaller is better
-------------------------------------------------------------------
Host OS 2p/0K Pipe AF UDP RPC/ TCP RPC/ TCP
ctxsw UNIX UDP TCP conn
--------- ------------- ----- ----- ---- ----- ----- ----- ----- ----
stp2-001 Linux 2.6.10- 8.270 38.6 24.3 61.6 48.5 45.9 76.6 74.6
stp2-001 Linux 2.6.10- 8.170 43.5 24.5 58.0 54.8 45.6 63.4 74.7
stp2-001 Linux 2.6.10- 2.740 50.6 29.9 40.3 48.3 59.8 75.1 101.
stp2-001 Linux 2.6.10- 8.140 46.6 29.7 57.6 48.8 45.5 72.0 74.4
stp2-001 Linux 2.6.10- 2.690 47.1 26.3 40.8 48.9 45.5 75.4 74.8
After TCP connection latency mean 73.8 std 0.55
*Local* Communication latencies in microseconds - smaller is better
-------------------------------------------------------------------
Host OS 2p/0K Pipe AF UDP RPC/ TCP RPC/ TCP
ctxsw UNIX UDP TCP conn
--------- ------------- ----- ----- ---- ----- ----- ----- ----- ----
stp2-001 Linux 2.6.10- 8.260 38.1 25.7 63.3 48.1 66.6 75.4 74.9
stp2-001 Linux 2.6.10- 8.090 46.2 26.0 63.4 55.5 45.9 63.6 73.5
stp2-001 Linux 2.6.10- 8.210 39.0 21.2 63.1 55.4 58.8 63.8 73.5
stp2-001 Linux 2.6.10- 2.850 46.5 26.0 64.8 54.6 45.5 74.0 73.6
stp2-001 Linux 2.6.10- 8.200 42.9 21.5 64.9 55.6 62.4 64.1 73.5
next prev parent reply other threads:[~2004-12-04 5:42 UTC|newest]
Thread overview: 21+ messages / expand[flat|nested] mbox.gz Atom feed top
[not found] <20041027092531.78fe438c@guest-251-240.pdx.osdl.net>
[not found] ` <200410291048.01955.michael.vittrup.larsen@ericsson.com>
2004-10-29 17:28 ` [PATCH] tcp: efficient port randomisation Stephen Hemminger
2004-11-01 9:58 ` Michael Vittrup Larsen
2004-11-01 17:20 ` Stephen Hemminger
2004-11-02 7:54 ` Michael Vittrup Larsen
2004-11-04 18:01 ` Stephen Hemminger
2004-11-05 10:03 ` Michael Vittrup Larsen
2004-11-17 23:30 ` [PATCH] tcp: efficient port randomisation (revised) Stephen Hemminger
2004-11-19 7:38 ` Michael Vittrup Larsen
2004-12-01 5:46 ` David S. Miller
[not found] ` <20041201152446.3a0d5ce3@dxpl.pdx.osdl.net>
[not found] ` <20041201204622.7b760400.davem@davemloft.net>
2004-12-02 21:49 ` Stephen Hemminger
2004-12-02 21:52 ` David S. Miller
2004-12-02 22:51 ` Stephen Hemminger
2004-12-02 23:01 ` Stephen Hemminger
2004-12-04 5:42 ` Stephen Hemminger [this message]
2004-12-06 8:18 ` Michael Vittrup Larsen
2004-12-06 17:42 ` [PATCH] tcp: efficient port randomisation (rev 3) Stephen Hemminger
2004-12-09 7:55 ` David S. Miller
2004-12-11 1:09 ` [PATCH] tcp: efficient port randomistion " Stephen Hemminger
2004-12-20 23:39 ` David S. Miller
2005-06-22 9:17 ` Michael Vittrup Larsen
2005-06-22 16:44 ` Stephen Hemminger
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=41B14E57.5080803@osdl.org \
--to=shemminger@osdl.org \
--cc=davem@davemloft.net \
--cc=michael.vittrup.larsen@ericsson.com \
--cc=netdev@oss.sgi.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).