bfifo limit explanation
Fernando Xavier <[email protected]> Thu, 30 Dec 2021 23:57:02 +0000
| Newsgroups | org.kernel.vger.lartc |
|---|---|
| Message-ID | <[email protected]> |
Hi there, I'm wondering If someone can clarify some pfifo-bfifo questions So, pfifo use packets and its limit defaults to txqueuelen and bfifo is txqueuelen multiplied by the interface mtu, default txqueuelen is 1000 on my machine But what if I play around with this and set a txqueuelen to a gigantic value like 10000 but set the bfifo limit to my bandwidth*delay product, upon congestion what will happen? I assume the bfifo will limit the queue to the value I want but will this interfere with the interface queue length? My plan is setting bfifo limit in my interface to the bdp assuming 100ms of delay So with an egress of 50mbit and assuming 100ms delay I have a bdp of 625000, if I set this as bfifo limit, this means that the queue can hold up to 100ms of data? Is this a correct thing to do? Does that make sense? Also, I noticed that in my machine (Debian 11, Kernel 5.10) I have access to pfifo_head_drop but there is no bfifo_head_drop, why is that? thanks.