Re: linear search of TCP control blocks on demultiplexing of an incoming IP datagram

Murali Bashyam <mbashyam@cisco.com> Wed, 17 April 2002 18:40 UTC

Message-ID: <3CBDC187.F1E4D47B@cisco.com>
Date: Wed, 17 Apr 2002 11:40:07 -0700
From: Murali Bashyam <mbashyam@cisco.com>
Organization: Cisco Systems Inc
X-Mailer: Mozilla 4.51C-CISCOENG [en] (X11; U; SunOS 5.6 sun4u)
X-Accept-Language: en
MIME-Version: 1.0
To: "Vishnepolsky, Oleg" <ovishnepolsky@doubleclick.net>
Cc: "'tcp-impl@grc.nasa.gov'" <tcp-impl@grc.nasa.gov>
Subject: Re: linear search of TCP control blocks on demultiplexing of an incoming IP datagram
References: <03E81ABDC8C2D31192C9009027D5B86E08ADB6CD@NYC-EX03>
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Sender: owner-tcp-impl@grc.nasa.gov
Precedence: bulk
Status: RO
Content-Length: 796
Lines: 20

"Vishnepolsky, Oleg" wrote:

> Which TCP implementations do a hash lookup ? I know that most BSD
> derivatives do a linear search, except for the freebsd stack which does a
> hash. What about MS stacks ? My stacks (OS/2, FlexOS) being BSD derivatives
> used linear search except I used a one control block cache lookup since most
> of the TCP packets come in "trains". I had a 70% hit ration in my
> experiments.

Actually the NetBSD implementation another BSD derivative also hashes the five
tuple, i'm sure
even the Linux implementation does.

Murali

>
> Is there a study/comparison of various TCP stacks re this and many other
> characteristics ? I am aware of external TCP fingerprinting  but that does
> not reveal the internals that much. Any help will be appreciated.
> Oleg Vishnepolsky

linear search of TCP control blocks on demultiple… Vishnepolsky, Oleg
Re: linear search of TCP control blocks on demult… Rick Jones
Re: linear search of TCP control blocks on demult… Luigi Rizzo
Re: linear search of TCP control blocks on demult… Murali Bashyam
Re: linear search of TCP control blocks on demult… David Borman
Re: linear search of TCP control blocks on demult… Alan Cox