After struggling with some strange effects with Bandwidth I have learned from Mikrotik Support that Flow Control is disabled in ROS on all interface.

Why it is needed. If you have links with different speed in your network the slower links cant forward Traffic not as fast as they arrive from the faster Link. This gives not only problem when the slower link is at full speed. Problems appear at lower saturation as traffic does not travel at equal spaced intervals. So a short burst of traffic can overload the slower link.

There comes Flow control. When the buffers of the device on one side of the slower link fills up it sends pause frames in the direction of the faster link to slow down traffic. If this device is a MT Router it ignores this pause frames and sends as fast as packets arrive and overflows the slower link.

We see this happening on Licensed links where the buffers are 128KByte way before the link is filled up. Packets are dropped and overall performance drops.

MT Support does not see Flow Control as a valuable Feature and asked my to ask this mailing list if there are others who need a working Flow Control. May be then there is a developer slot to just make the knob to enable it.

Anyone else need Flow Control ?

I think it can be very helpfull as we are sorting out this problem namely once there is reduction of the speed between link capacities - especially Gbit ethernet to Radio which is lets say 200Mbit. This singificantly reduce thruput..

I think it can be very helpfull as we are sorting out this problem namely once there is reduction of the speed between link capacities - especially Gbit ethernet to Radio which is lets say 200Mbit. This singificantly reduce thruput..

me too struggling with it...

So, we have Flow Control in the new RouterOS 6.11. Let's try it now!

I have a similar situation here:

RB1000 connected:
eth6 WAN link (100mbit FD bandwidth available) but connected to carrier's CPE at GigE speed.
eth1 1Gbps to Media Converter (gig speed)

Media converter connected at GigE speed to UBNT Thoughswitch 5
From the TS-5 I have a powerbridge M5 from UBNT with 100mbit port.

If I download something via the PB-5 I got 20mbit, if I use a laptop linked at 100mbit to the switch, I got 100mbit.

How can I resolve?I enabled theflow control but the situation remained unchanged.

So, we have Flow Control in the new RouterOS 6.11. Let's try it now!

Looking at the other reports on 6.11 we'll wait for 6.12 with testing ...
There is another thread reporting some issues where pfifo queuing introduces speed problems. sfp queuing solves this. In the past we switched away from sfp as it slowed down single users. pfifo helped. Another thread's user tested Flow Control and did not see a difference.

So it is still not clear what is going on. We just installed a relay tower and see >120MBit UDP and <60MBit tcp to the end user. One link to the tower, one SA-sector with one SXT Lite connected. All nv2 dual channel/40MHz.

So, is the flow control a good thing or not? I've found this negative old article: http://virtualthreads.blogspot.sk/2006/ ... ntrol.html
What do you think?

So, is the flow control a good thing or not? I've found this negative old article: http://virtualthreads.blogspot.sk/2006/ ... ntrol.html
What do you think?

It depends. We use it at links where speed changes. E.g. 100M to 1G or Ethernet to wireless. There it helps to reduce the impact of bursty traffic.

So, is the flow control a good thing or not? I've found this negative old article: http://virtualthreads.blogspot.sk/2006/ ... ntrol.html
What do you think?

It depends. We use it at links where speed changes. E.g. 100M to 1G or Ethernet to wireless. There it helps to reduce the impact of bursty traffic.

My friend, you always seem to be able to give me good advice.

We are improving our network lately a lot with OSPF and PPP tunnels to clients and at the same time upgrading hardware and improving wireless links.

But we get a 'feeling' that the better we make the network, the more complaints we get from people saying 'the internet is sooo slow...!'
We are breaking our heads over this since MT's bandwidth tests from MT cpe's always show very good results towards our end router.
Even tests done from a laptop show good, (not the same as very good) results.
But www.speedtest.net or other tcp downloads are mostly only 50% of the same speeds....

Even our CCR that connects to the Cisco of our provider with the 300/300Mb line is not able to produce single downloads reaching even near this limits. If I run a single tcp download from my powerfull PC that has an 'unlimited' queue set I only get 100, 150Mbps in at best. Sometimes way less. Only if we open several multi stream downloads with a multi download stream program we in the end reach 280-290Mbps downloads...... we are a bit puzzled by why this is as it is....

So today we discovered some threads about the 'traffic flow' which in all our network is by default 'off'. (All MT network)

Our network has indeed a mixture of fast Ethernet and gigabit Ethernet in any 'route' traffic flows to reach remote client. Many routers that make wireless backhauls have on one end either a Gb Ethernet or a fast Ethernet where not always the wireless links themselves can produce 100Mbps (let alone Giga!). Like the links between AP and client for instance.
Plus the fact Ethernet is full duplex where wifi is simulated level 3 duplex at best but 98% is just simplex.

So we where hoping this traffic flow switch might bring us salvation..... until I've red someone that said it will kill VoIP!
We don't have a lot of VoIP clients (Is Skype VoIP?? If so then we do have al lot more...) but these that have are also the better paying ones. We wouldn't want to waste these clients....

Any advice?

So, is the flow control a good thing or not? I've found this negative old article: http://virtualthreads.blogspot.sk/2006/ ... ntrol.html
What do you think?

It depends. We use it at links where speed changes. E.g. 100M to 1G or Ethernet to wireless. There it helps to reduce the impact of bursty traffic.

My friend, you always seem to be able to give me good advice.

We are improving our network lately a lot with OSPF and PPP tunnels to clients and at the same time upgrading hardware and improving wireless links.

But we get a 'feeling' that the better we make the network, the more complaints we get from people saying 'the internet is sooo slow...!'
We are breaking our heads over this since MT's bandwidth tests from MT cpe's always show very good results towards our end router.
Even tests done from a laptop show good, (not the same as very good) results.
But http://www.speedtest.net or other tcp downloads are mostly only 50% of the same speeds....

Even our CCR that connects to the Cisco of our provider with the 300/300Mb line is not able to produce single downloads reaching even near this limits. If I run a single tcp download from my powerfull PC that has an 'unlimited' queue set I only get 100, 150Mbps in at best. Sometimes way less. Only if we open several multi stream downloads with a multi download stream program we in the end reach 280-290Mbps downloads...... we are a bit puzzled by why this is as it is....

So today we discovered some threads about the 'traffic flow' which in all our network is by default 'off'. (All MT network)

Our network has indeed a mixture of fast Ethernet and gigabit Ethernet in any 'route' traffic flows to reach remote client. Many routers that make wireless backhauls have on one end either a Gb Ethernet or a fast Ethernet where not always the wireless links themselves can produce 100Mbps (let alone Giga!). Like the links between AP and client for instance.
Plus the fact Ethernet is full duplex where wifi is simulated level 3 duplex at best but 98% is just simplex.

So we where hoping this traffic flow switch might bring us salvation..... until I've red someone that said it will kill VoIP!
We don't have a lot of VoIP clients (Is Skype VoIP?? If so then we do have al lot more...) but these that have are also the better paying ones. We wouldn't want to waste these clients....

Any advice?

Best would be to eliminate this 100M devices out of your network. Second best is to enable flow control. There it depends on the devices how they do flow control to make a smooth slow down of traffic at the overflowed interface.

Best would be to eliminate this 100M devices out of your network. Second best is to enable flow control. There it depends on the devices how they do flow control to make a smooth slow down of traffic at the overflowed interface.

Hmm, but what do you refer to as "network"? Does that include the client's wifi router? That is almost by default 100M?
Or does it included the 700 SXT's 'n' we have? They are all 100M.

3/4 of our AP's are OmniTiks. They have 100M connections only.
We also have still several rb750's in place for power supply with line check. Up to now MT doesn't even have a Gig replacement for these......

So making the full network to Gigabit is hard if not impossible and depending to what we see as beeing 'the network'?