Jumbo frames , Catalyst 6XXX , gigabit links

[rrsstefano]

Scenario :

SUN NFS Server ------- Catalyst 65XX -------- Client NFS SGI

-- Client and server are on the same vlan

-- Client NFS SGI support jumbo frame

-- SUN Server doesn't support jumbo frame

-- Catalyst configured to support jumbo on client side

Because the PMTUD negotiation, client use the default ethernet MSS (1460 )

with bad performance......

The performance between two SGI clients using jumbo is much better regarding gigabit utilization

Is possible to bypass this problem?? disabling PMTUD ? or using the msfc2 on Catalyst , making a layer3 routing , so that the Catalyst router engine can make fragmentation??

I think that the layer 3 solution only shift the problem on Catalyst ....

All suggestion are greatly accepted!!!!!

Thanks

[jbohla]

I agree with you that the layer 3 is a good solution. But do consider the impact before making the dcision. There are several issues that make IP fragmentation undesirable. There is a small increase in CPU and memory overhead to fragment an IP datagram. Another fragmentation issue involves handling dropped fragments. If one fragment of an IP datagram is dropped, then the entire original IP datagram must be resent, and it will also be fragmented. For details refer to:

http://www.cisco.com/en/US/tech/tk648/tk369/technologies_white_paper09186a00800d6979.shtml#subfirstone

[rrsstefano]

Effectively , using fragmentation seems to be, at the moment, the unique idea...but, as you suggested, it must be studied carefully.

I'm not an expert about SUN NFS and so on, but is really impossible to use jumbo frames on SUN???

Do you have experience about it?

Regards

Stefano