RE: calculate correct MTU size for jumbo frames?

From: Matthew Zito <mzito_at_gridapp.com>
Date: Sun, 9 Nov 2008 00:56:21 -0500
Message-ID: <C0A5E31718FC064A91E9FD7BE2F081B1D520D3@exchange.gridapp.com>

It's too late here, I neglected to include the URL - http://sd.wareonearth.com/~phil/net/overhead/

Also, another way to tell is to use a tool like tcpdump/wireshark and watch the cache fusion traffic and see if the frames are being fragmented.

Finally, this also doesn't apply if you're using something like LACP or 802.1q, both of which will increase the frame overhead due to additional metadata.

Thanks,
Matt

-----Original Message-----

From: oracle-l-bounce_at_freelists.org on behalf of Matthew Zito Sent: Sun 11/9/2008 12:21 AM
To: martin.a.berger_at_gmail.com; ORACLE-L
Subject: RE: calculate correct MTU size for jumbo frames?  

9000 is the most commonly used value, and as a nice round value, less opportunity for misconfiguration. Some old switches could only do 4400 bytes, but at this point, you should just set it at 9k with the rest of the world ;)

If you did want to calculate, you just need to sort out ethernet overhead + ip overhead + tcp overhead, which based on this site (first hit on ethernet overhead on google) is 44 bytes.

Thanks,
Matt

--

http://www.freelists.org/webpage/oracle-l Received on Sat Nov 08 2008 - 23:56:21 CST

Original text of this message