[Bioclusters] Jumbo frames with 100 Mb?

[Joe Landman]

On Thu, 2002-05-02 at 13:50, Vsevolod Ilyushchenko wrote:
> Hi,
> 
> Sanity check:
> 
> Am I correct that enabling jumbo frames on a Gb card makes sense only if
> there is another Gb card on the other end? 

Yes.  The packets will be fragmented up to the maximum size (the MTU). 
For standard 10/100 ethernet, this is 1500 bytes.  Jumbo frames are much
larger.

> That is, if the master has the
> GigE card, but all the slaves have only 100 Mb cards, jumbo frames will not
> help or even screw everything up, right?

I believe there is an initial negotiation phase during connection with
the hardware (card <-> switch ) that builds a greatest common feature
set, so if you plug in a jumbo frames capable GigE card into a 10/100,
it will negotiate down to the best effort of the slower interconnect
(and no-jumbo frames would be supported).

> Also, do I have to do anyting to the switch to enable jumbo frame support?

You need to check the switch specs for it.  There are a fair number of
cards out there advertised as GigE cards, but without Jumbo frames
support.  I wouldnt be suprised if the switchs have the same gotchas.

> 
> Thanks,
> Simon
> -- 
> Simon (Vsevolod ILyushchenko)   simonf@cshl.edu   
> http://www.simonf.com          simonf@simonf.com 
> 
> The central message of Buddhism is not 
> "every man for himself"!
> 				("The Fish Called Wanda")
> _______________________________________________
> Bioclusters maillist  -  Bioclusters@bioinformatics.org
> http://bioinformatics.org/mailman/listinfo/bioclusters
-- 

Joseph Landman, Ph.D.
Senior Scientist,
MSC Software High Performance Computing
email		: joe.landman@mscsoftware.com
messaging	: page_joe@mschpc.dtw.macsch.com
Main office	: +1 248 208 3312
Cell phone	: +1 734 612 4615
Fax		: +1 714 784 3774