Page 1 of 1

flow control 8..9..10?

Posted: Fri Apr 21, 2017 10:37 pm
by cyberbroadband
i may have some underpowered mikrotik. i am not sure yet.

symptom - radios at the end of the line are not giving individual customers much speed - but the backhaul appears to be performing.
an end user customer on a 10 x 10 plan might be seeing 7 x 2.

there are 5 tower hops to get there. very rural area. not many options.
this is a relatively new build.

tower a -> tower b -> tower c -> tower d -> tower e

tower a and tower b are preexisting, tower a has always been "busy", tower b not so much - 3 to 10 meg average. tower c, d, and e are new.

business customer on tower e isn't seeing good upload. they need it. (they want to run a terminal server and let some remote offices connect in)

question -

on tower a, we have two netonix switchs, both 16 port.
they are connected with a short patch cable, linking 1g to 1g.
since this is wired, flow control recommended or not?

i have flow control enabled on everything all the way down the line.

thank you in advance.

Re: flow control 8..9..10?

Posted: Fri Apr 21, 2017 11:01 pm
by sirhc
In large flat networks unless you really know how flow control works and understand how to monitor the Pause Frames and determine what works best I would turn it off.


I have a video on Youtube you can go watch but my network is based on a router at every tower based on OSPF.

https://www.youtube.com/watch?v=8JvBEAD4MFM

Re: flow control 8..9..10?

Posted: Fri Apr 21, 2017 11:09 pm
by cyberbroadband
Thanks - I was trolling about this several months ago (might be eight months ago now) ; think I watched most of that and read about every thread there was on it...

Just wasn't sure how to handle a switch to switch situation. Seems most people flow control on everything except their backhaul if I remember correctly (assuming their backhaul is a gig ethernet port)

Re: flow control 8..9..10?

Posted: Sat Apr 22, 2017 12:21 pm
by mike99
It's the opposite, flow-control is most usefull on ptp link since it block the port and you would stop traffic for every device on the PTMP the time one device lower is buffer, introducing latency spike and broking TCP that need device to drop paquet to adjust the flow not keep those on a other device that buffer that is not full.