PCQ limit and Total Limit parameters explanation

RouterOS General
1

Hi all,

There are a dozens of unanswered questions regarding PCQ parameters on the Forum during the years (I’ve spent two days browsing forum and Internet) - I can list these topics here if someone is interested (the questions were not asked by noobs but an experienced networking people).
I agree that using various resources (Megis presentation, MUM archives, Wiki, Manual, Forum etc.) and with a little bit of experimenting you can come up to the values that fits specific solution, but I don’t think you can truly understand what is all about it.

Can someone please clear this out and explain “pcq-limit” and “pck-total-limit” values in more detail (especially when “pcq-rate” option is not set).
If this is a secret because PCQ is Mikrotik proprietary - please say so and we probably will not ask so many questions about it.

As a starting point we can look at Megis presentation slide 29 from 2008 (attached) and comment on the values (i’m not even sure if the units in this settings are in KiB because there is some strange multiplication in “Total Limit” formula - maube this was different in old RouterOS versions). Also, I’m not sure why the packets are counted as 2200 bytes.
There is no mention about any timers involved - but it seems there must be some for this to function properly.

It would be of an additional value if someone provide us with the efficient solution for testing this kind of settings and PCQ behaviour in the lab environment.

Kind regards,
Moky
Megis_presentation_2008_slide.png

2

Nobody has an explanation even for this Megis’s slide? :confused:

3

So what’s the question?

According to the Manual:

pcq-limit (number) : queue size of single sub-stream (in KiB)
pcq-total-limit (number) : maximum amount of queued data in all sub-streams (in KiB)

About 2200 bytes of RAM - it can be some error, because earlier pcq-*-limit were mentioned to be measured in packets, not in KiBs, but it appeared that it’s still in KiB, and the manual was edited. Probably Janis M. was counting limits in packets, following that stale data.