When should we use switches in cut-through mode?

[mad_mavula]

I know switches in cut-through mode don't check for errors in frames, but I need to know why exactly we don't need to check frames for errors and why should we use cut-through mode.

[balaji.bandi]

Depends on the model and depends on the context of discussion :

below give you high level how the switch work on cut through :

https://www.networkacademy.io/ccna/ethernet/store-and-forward-vs-cut-through-switching

cisco Moving all Cat 9K models here is good discussion help you :

https://community.cisco.com/t5/switching/catalyst-9k-switching-modes-cut-through-vs-store-and-forward/td-p/4442520

Hope this information help you.

[mad_mavula]

Yeah I had the idea of cut through mode. But I need the exact answer on why do we use that. Is it just about fast forwarding or is there anything else? are there any occasions  that we shouldn't use cut-through mode?

[Joseph W. Doherty · Accepted solution]

It's for faster forwarding.  It reduces switch latency queuing the frame (which increases with larger frame [especially, for example, for "jumbo" Ethernet max sized frames]).

BTW, even with cut-through, the "old" hub/repeaters have less latency per frame (for same bandwidth).

When not to use it?  When you don't want to forward possibly already corrupted frames.  Or when you don't have real need for it and you don't want to pay for the increased device cost, for devices that support it.

Also BTW, I recall "original" cut-through switches were to "compete" with the hub/repeaters they were replacing, providing much reduced latency for large frames (i.e. closer to a hub/repeater's performance), especially at 10 Mbps.  When switches moved to gig bandwidths, frame latencies, being 100x less, i.e. gig vs. 10 Mbps, switch frame queuing latencies were generally considered not significant.  With both the rise of low latency demanding applications (e.g. on-line trading), and "jumbo" Ethernet, cut-through technology has re-appeared.