Re: Iperf max b/w

To: Masaki Hirabaru <masaki --at-- crl.go.jp>
Subject: Re: Iperf max b/w
From: Cameron Harr <charr --at-- lnxi.com>
Date: Fri, 10 Oct 2003 17:59:59 -0600
Cc: iperf-users --at-- dast.nlanr.net
Content-transfer-encoding: 7bit
Content-type: text/plain; charset=us-ascii; format=flowed
In-reply-to: <20031011.085107.60852039.masaki --at-- crl.go.jp>
References: <3F86BF7E.6000501 --at-- lnxi.com> <20031011.085107.60852039.masaki --at-- crl.go.jp>
Reply-to: iperf-users --at-- dast.nlanr.net
Sender: owner-iperf-users --at-- dast.nlanr.net
User-agent: Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.4) Gecko/20030624 Netscape/7.1 (ax)

Masaki Hirabaru wrote:

> If we use regular MTU 1500B on GbE and running TCP with timestamp
> option (Linux default?), the maximum TCP throughput will be 941
> Mbps. It's the line rate of 1 Gbps.
> 
> I checked iperf source code and found that the number iperf gives us
> is a rate at application layer, i.e. hidden overhead excluded from the
> number will be TCP header and options, IP header and options, and
> ethernet gap, preamble, header, and trailer. In my case, these are 32
> (TCP), 20 (IP) and 38 (ethernet) bytes long.
> 
> With the same calculation, UDP will not go over 957 Mbps with MTU
> 1500B on GbE. If you got those numbers, I think your equipment can be
> considered well tuned under that condition.
> 
> Masaki

Thank you. I haven't tried UDP, but the TCP seems to be maxed out at 
line speed.

-- 
Cameron Harr
Systems Engineer
Linux Networx Inc
http://www.lnxi.com

References:
- Iperf max b/w
  - From: Cameron Harr
- Re: Iperf max b/w
  - From: Masaki Hirabaru

Prev by Date: Re: Iperf max b/w
Next by Date: Re: Iperf max b/w
Previous by thread: Re: Iperf max b/w
Next by thread: RE: Iperf max b/w

Other Mailing lists | Author Index | Date Index | Subject Index | Thread Index