Someone writing a feature request to add jumbo frame support to the TCP/IP stack for Mac OS X is looking around for real world 'before and after' performance data. People seem happy doing jumbo frames in switch-to-switch configurations with homogenous network hardware but has anyone enabled this out to client hosts in one of their clusters? I've only done this as a way to get familiar with the management interfaces on various bits of hardware and have never left the settings enabled for any length of time or in a real production environment. Any stories or data to share? -Chris