Editing Cut-through switching (section)

==Use in Ethernet==

When cut-through switching is used in [[Ethernet]] the switch is not able to verify the integrity of an incoming frame before forwarding it.

The technology was developed by [[Kalpana (company)|Kalpana]], the company that introduced the first [[Ethernet switch]].<ref>{{cite web|url=http://newsroom.cisco.com/dlls/1994/corp_102494.html |title=Cisco to Acquire Kalpana, Leading Ethernet Switching Company |archiveurl=https://web.archive.org/web/20100207131818/http://newsroom.cisco.com/dlls/1994/corp_102494.html |archivedate=2010-02-07 |publisher=Cisco Systems, Inc. |url-status=dead }}</ref>

The primary advantage of cut-through [[Ethernet switch]]es, compared to store-and-forward Ethernet switches, is lower latency.<ref name="cisco-cut-through" >
Cisco.
https://www.cisco.com/c/en/us/products/collateral/switches/nexus-5020-switch/white_paper_c11-465436.html "Cut-Through and Store-and-Forward Ethernet Switching for Low-Latency Environments"].
</ref>
Cut-through Ethernet switches can support an end-to-end [[network delay]] latency of about ten microseconds.
End-to-end application latencies below 3 microseconds require specialized hardware such as [[InfiniBand]].<ref name="cisco-cut-through" />

A cut-through switch will forward corrupted frames, whereas a store and forward switch will drop them.<ref>{{cite web |url=http://www.cisco.com/en/US/prod/collateral/switches/ps9441/ps9670/white_paper_c11-465436.html |title=Cut-Through and Store-and-Forward Ethernet Switching for Low-Latency Environments |publisher=Cisco |accessdate=2011-11-10}}</ref> '''Fragment free'''{{anchor|Fragment free}} is a variation on cut-through switching that partially addresses this problem by assuring that collision fragments are not forwarded. Fragment free will hold the frame until the first 64 bytes are read from the source to detect a collision before forwarding. This is only useful if there is a chance of a collision on the source port.<ref>{{cite web|url=http://support.intel.com/support/express/switches/sb/cs-014410.htm |title=Switches - What Are Forwarding Modes and How Do They Work? |accessdate=2011-08-13 |url-status=dead |archiveurl=https://web.archive.org/web/20140419110752/http://www.intel.com/support/express/switches/sb/cs-014410.htm |archivedate=2014-04-19 }}</ref>

The theory here is that frames that are damaged by collisions are often shorter than the minimum valid Ethernet frame size of 64 bytes. With a fragment-free buffer, the first 64 bytes of each frame update the source MAC and port if necessary, provide the destination MAC, and allow forwarding the frame. If the frame is less than 64 bytes, it is discarded. Frames that are smaller than 64 bytes are called runts; this is why fragment-free switching is sometimes called "runt-less" switching. Because the switch only ever buffers 64 bytes of each frame, fragment-free is a faster mode than store-and-forward, but there still exists a risk of forwarding bad frames.<ref>{{cite web |url=http://www.thesupportsystem.co.uk/2010/04/09/switching-store-and-forward-cut-through-and-fragment-free/ |title=Switching – Store and forward, Cut-through and Fragment free |accessdate=2013-11-11 |url-status=dead |archiveurl=https://web.archive.org/web/20131111215547/http://www.thesupportsystem.co.uk/2010/04/09/switching-store-and-forward-cut-through-and-fragment-free/ |archivedate=2013-11-11 }}</ref>

There are certain scenarios that force a cut-through Ethernet switch to buffer the entire frame, acting like a store-and-forward Ethernet switch for that frame:
* Speed: When the outgoing port is faster than the incoming port, the switch must buffer the entire frame received from the lower-speed port before the switch can start transmitting that frame out the high-speed port, to prevent underrun. (When the outgoing port is slower than the incoming port, the switch can perform cut-through switching and start transmitting that frame before it is entirely received, although it must still buffer most of the frame).
* Congestion: When a cut-through switch decides a frame from one incoming port needs to go out through an outgoing port, but that outgoing port is already busy sending a frame from a second incoming port, the switch must buffer some or all of the frame from the first incoming port.<ref name="cisco-cut-through" />