Unsolved
43 Posts
0
2209
Cut-through mode configuration
Hello,
i try to wrap my head around how DELL networking switches handle cut-through switching. Officially their ASIC features this mode, but i seem to can't find any option how to configure it.
Or is it buried implicitely in some other config mode realted to packet forwarding?
It would affect switches of the DELL networking S*-class...
If someone has a strong opinion about the use of cut-through in modern low-latency switches in general, this would also be highly welcome.
Best
DELL-Josh Cr
Moderator
Moderator
•
8.7K Posts
0
September 16th, 2019 09:00
Hi,
Which model switch are you trying to configure? Which OS version are you using?
HifDelCo
43 Posts
0
September 17th, 2019 10:00
Hello Josh,
thank you for willing to help!
The device i currently use is a S4810P (OS 9.14(1.2)), and the possible future device would be a S5048-ON or S5248-ON, then with the latest and greatest release.
So far i lived in the perception c-t is a sort of default mode, switched on in the ASIC, and there is no need to configure it, because the ASIC is taking care of it.
But then i found that some other OSes provide a configuration option for controlling it...
I also wonder in general, whether technical progress may have shifted the situation the cut-through mode is designed to tackle, and how that would affect it's configuration in specific use cases...
Best regards,
H
DELL-Josh Cr
Moderator
Moderator
•
8.7K Posts
0
September 18th, 2019 10:00
I do not see a way to configure it on these switch OSes. Some of the Z series switches and the S6100 have a custom OS version that supports it.
HifDelCo
43 Posts
0
September 19th, 2019 01:00
Hello Josh,
It's weired, because i'm quite shure i observed some configuration snippets for a S4810P in a troubleshoot discussion about a different issue on the forum, and it had a line similar to "no cut-through", and it mattered in that discussion.
But i can't track it down anymore.
It left me with the idea it can at least be deactivated.
I'm interested in it because i don't exactly know how the ASIC would deal with situations where congestions or underruns (e.g. due to application behaviour) or other imperfections in the network occur: would the ASIC switch to store-and-forward altogether? And would that happen in a fashion that the latency values effectively stay predictible?
Best,
H