Had an interesting issue at work. There is an application that sends really huge frames > 1500bytes. Our environment spans quite a few hops to the receiver of this app's data. There is a mixture of store/forward and nexus 3ks (cut through) and Aristas. Analyzing the span capture off the Arista interface leading to the receiver we noticed the huge packets but where left scratching our heads....How is a jumbo frame being sent throughout our network if the MTU is 1500. After some digging through data sheets we came to the conclusion that even though a switch is not configured on its interface to support jumbo, if the platform itself can support it it will still process the frame and forward it.
The takeaway is that cut-through switches only look at the destination frame address. Therefore it doesnt care how big the packet size is and so jumbo frames are sent on egress interfaces that are set with 1500 bytes by default
Store/Forward switches such as a 6509-e with a 720 card does support jumbo frames. Even though its not configured the mere fact that its supported means it will be forwarded.
Do my fellow engineers on this thread agree or disagree?
No comments:
Post a Comment