hi all.
Is it possibile to improve performances of a sector with 50 users in 5GHz,
only enabling Nv2 only on access point,
leaving cpe’s at ros 3.30-4.10 ?
thanks
No, the CPEs also need to support nv2 so they’ll need to be upgraded to the latest 5.x version as well.
And what would happen if we have only some (1-4) other manufacturer’s CPE’s in the AP network where all the rest would be MT-ros5rc1 nv2 enabled CPE’s?
Are these ´strange´ CPE’s thrown out of the AP? Or will only their performance degrade? Or will the whole CPE network not work?
Hi, what I see nv2 on MtMP links no performance improvment. We tested 1 AP (A-Band) with 12 clients which was running for 2 years with nstreme. We are able to get around 12-16 Mbits to the clients.
After upgrading to 5.0rc1 we swichted to nv2 and got only 8-10 MBit/s., bad ping times - 25 - 200 ms - but stable connections to all clients.
On a PtP link we get very good speed and stable ping times with nv2.
I dont belive you will get better performace out of your AP with nv2.
Well, I think the nv2 is basically designed to give improvement on busy AP’s. Because each client should be given now a dedicated timeslot to send and recieve instead of waiting for free airtime with AP. But as always, some further config settings are needed to have it working like it should.
I also think nv2 will work better when more ‘hidden nodes’ are present and many busy clients are demanding airtime over AP’s that only have some short distant clients with little traffic demand.
I also wonder how your speed test to the clients is done. If you just pick a client and see what the maximum througput can be it could well be that with nv2 the result is a bit lower. Since with nv2 the client doesn’t get all airtime, even if it is the only CPE demanding it. Slots are kept available for the other clients. With the old nstreme I don’t think that is the case. Here the client gets all the airtime needed if the other don’t ask for any.
So the improvemen to nv2 will be when several clients are demanding airtime at the same time. In the old setup with nstreme (or 801.11 rts/cts system) each client has to wait untill airtime is freed for him and one client can sort of ‘claim’ all airtime when he has a big traffic flow over the link. Because all clients are still sending time request to the AP the AP has to consume this all and when there are also many hidden nodes things become chaotic for the AP.
Here now nv2 should give an improvement towards all clients if they all are demanding airtime because they all want to transfer high amounts of traffic. The AP just give each client a ‘time slot’ and clients just wait for their turn. Now, with some other improvement in the nv2 mode you can even give some priority to some more importain clients or higher priority traffic so this basically should improve overal performance if settings are done right.
So to test both systems against eachother I think you need to run a bandwith test over several clients at the same time and then try both systems. If the nv2 settings are also done like they should I am convinced it will bring better speed and lower latency to EACH client under heavy usagae conditions for ALL clients.
It still leave my question unanswered. What happens when 3rd party CPE’s are in the AP’s network?
Since nv2 is a MT protocol 3rd party device don’t work with it.
But what are the possible results if they are still around? Are they omitted any airtime? Or do they create ‘collision’s’ in the AP-network?
I have some AP’s with 20+ clients in very congested ‘air spectrum’ and several ‘hidden nodes’. But each of these AP’s still have 4 or 5 nanostation in them. Do I first need to replace them before I am going to implement the nv2 or not?
I am dying to try the nv2 system but can’t affored to disconnect some clients nor to replace many antennas at once.
You can only use mikrotik devices. UBNT is not ompatible
- nanostation cant connect to ap with nv2 enabled
- nv2 is not yet production grade (no security, …)
- A station searching for an AP should not disturb the nv2 ap
Dont mix MT and UBNT on wireless. Each has its own special protocol.
You can only use plain 802.11 when mixing them.
- So they are de-assosiated the moment I swith the AP to nv2? Ok, makes sense.
- Well, it is on ros5rc which makes it a ‘near production’ grade. There are many bugs in the ros5.x to be solved yet. Probably the first ‘production ready’ Ros5.x will still have issues. But in an all MT environment with only a basic AP/CPE-client network I think it is already usable. It is more the ‘odd’ configs or the ‘3rd party hardware’ that still has lots of issues. I don’t use security on my networks. I have no need for it and keeps the network simple.
If we have to wait untill all bugs are solved from ROS before we can call it ‘production grade’ we always have to wait until MT actually abondonce the family. Look at 2.x. There will still bugs around when 3.x came around. 3.30 is stable but still some minor bugs around. 4.11 still has several bugs waiting for solution. So when will ros5 be ‘production grade’?
It all depends what you want from ROS and what your hardware and network setup is. I presume that MT will work to make their own hardwork perform best before 3rd party stuff comes in.
I made the mistake two years ago to go with UBNT to safe some bugs then, but I am now more and more coinvinced that an all MT environment works best if you want a good and simple performing network. - Ok.
My networks are already mixed. It is all about migration into the new nv2. I am busy converting my networks back to MT but 20% of my clients have UBNT so that won’t happen overnight.
I considered ubnt as MT did not show progress with 11n for a Long time.
But my testbench with ubnt did not show good results. The same behavior keeps me
away from using nv2 in production until now. I test every new version but still
wait until the rough edges are sorted out.
In the meantime we deploy RBs which are capable of running it.
We are seeing very similar results on early tests. WDS bridged clients, ookla speedtest server (on our network). Mtik bandwidth tests get me 7M or so on both UDP and TCP (20 connections).
Yet again, how do you test? Run a bandwidth test on several CPS’s at the same time and see if that then gives any differences.
The improvement on PtP links is a result of some secondary improvements of the protocol. The real improvement is to be found on busy AP’s. Like MT already ask results for from the field.
It is the same as in the airspace above a big airport where all planes want to go and leave to destination. Off course most pilots can speak to each other over the radio and negotiate who has first right to enter and leave certain runways. But some might not be heard and most pilots have only a slight idea how many planes there really are. So when it becomes too busy an airport air controller would make some big improvements. He gives each plane that want to approach his airport a slot for entering and leaving and thus the amount of traffic can increase considerable while at the same time collisions can be avoided… Now it doesn’t matter anymore if some planes cannot see another one. Evert plane gets its turn and thus every plane is transited faster.
Some extra features of the controller (protocol) is that now the Airforce one gets better handling then some cargo buster full with UBNT radios!
That some planes can reach their international destination a bit faster after leaving the airport is merely because the controller gave more specific directions to reach it and is only a bonus of the new controller system.
But the aim is to help the congested area at the port (read “AP”.)
Now, running a speed test on a single plane wont tell us a lot about the maximum speed he can use to reach his runway.
But measuring several planes approach times and compare that to the pre-controller era would give us a better idea in how effective the new controller is.
I hope this little story might be appreciated! 
Hi all, i am using mikrotik since Version 1.xx and have deployed 100s of APs. As i can See nv2 is not ready for production network.
We tested in production network for arround 4 days.
Today we switched back to nstreme at that AP and throughput has nearly doppeld.
The Next worst behavior is latency of nv2 and throughput feels tenacious.
I Hope the Mikrotik Guys keep going their good work and we will get better performance of nv2 in a few .
I will post a graph tomorrow.
See you at mum usa -
here is a screenshot from our test AP. real traffic from users
