Community discussions

MUM Europe 2020
 
xbaha
Member Candidate
Member Candidate
Topic Author
Posts: 111
Joined: Fri Mar 09, 2007 12:16 pm

POINT TO MULTIPOINT ISSUE

Sun May 11, 2008 3:53 am

hi,
i have my main AP as AP Bridge, WDS Enabled, and there is 2 clients as WDS Station connected to the main AP, Say, Client A and Client B.
i am using WDS to be able to bridge the trafic between ether1 and wlan.

the problem is:

when i do bandwidth test between Client A, and the main AP, i notice that the ping between the AP and Client B rises, when i stop, it gets back to normal.
the same with the opposit, when do bandwidth test between client B and main AP, the ping between AP and client A rises.

AP is RB600, clients are 133.
when i do bandwidth test, the CPU utilization on the RB600 is max 20%.

I dont understand what is the relation?? is it WDS ? there is no other settings.
your input on this matter will really help.

Thanks all for any comments.
 
0ldman
Forum Guru
Forum Guru
Posts: 1446
Joined: Thu Jul 27, 2006 5:01 am

Re: POINT TO MULTIPOINT ISSUE

Sun May 11, 2008 7:03 am

btest is harder on the system than standard TCP traffic.

Another thing, WDS gets kind of weird with more than one device in the bridge. Also, set Queues to prevent any one machine from using all the available bandwidth, QoS control, etc...
 
xbaha
Member Candidate
Member Candidate
Topic Author
Posts: 111
Joined: Fri Mar 09, 2007 12:16 pm

Re: POINT TO MULTIPOINT ISSUE

Sun May 11, 2008 4:19 pm

i want to know a logical explination what makes the other link unstable, i dont want to build rules to make it stable, it should be stalbe without any rules!!
 
User avatar
Belyivulk
Member Candidate
Member Candidate
Posts: 285
Joined: Mon Mar 06, 2006 10:53 pm
Location: Whangarei, New Zealand
Contact:

Re: POINT TO MULTIPOINT ISSUE

Mon May 12, 2008 8:51 am

Sounds to me like when BT'ing your 'spamming' the channel, and utilising all available bandwidth for the one customer. That would be my guess.

Limit the BT to 2mbit or something and see if the problem continues
 
User avatar
jwcn
Forum Guru
Forum Guru
Posts: 1501
Joined: Sun Aug 27, 2006 6:49 am
Location: Maryland, USA
Contact:

Re: POINT TO MULTIPOINT ISSUE

Mon May 12, 2008 3:57 pm

it should be stalbe without any rules!!
Wrong.
 
User avatar
JJCinAZ
Member
Member
Posts: 473
Joined: Fri Oct 22, 2004 8:03 am
Location: Tucson, AZ
Contact:

Re: POINT TO MULTIPOINT ISSUE

Mon May 12, 2008 5:04 pm

802.11 is not a fixed TDM system. If one client is using a lot of airtime, the others have to wait. WDS is not a factor here. Using nstream with polling helps.
 
xbaha
Member Candidate
Member Candidate
Topic Author
Posts: 111
Joined: Fri Mar 09, 2007 12:16 pm

Re: POINT TO MULTIPOINT ISSUE

Mon May 12, 2008 6:59 pm

what do you mean by air time???
how much bandwidth can an empty channel give me ?
 
User avatar
JJCinAZ
Member
Member
Posts: 473
Joined: Fri Oct 22, 2004 8:03 am
Location: Tucson, AZ
Contact:

Re: POINT TO MULTIPOINT ISSUE

Mon May 12, 2008 7:20 pm

802.11 uses data frames and control frames. Control frames are transmitted using the Basic rate and the data frames are transmitted using the Supported Rates. This means that even if the link is "perfect" and able to do 54Mbps modulation at all times, you can't send 54Mbps through it because some of the frames are moved at 11Mbps. Further, there is frame overhead on the air, so you have further reductions in possible throughput.

Now, assume everything moves at 54Mbps, you can calculate how long it will take to transmit a 1538 frame (not everything sent will be that size mind you, so your overhead is always changing) on the air, not including interframe gaps and time to wait for clear air (802.11 uses listen and send unless you have NStream enabled with disable CDMA). That time value will be the MINIMUM possible latency for the second client in your test because the second client has to wait for the first to clear the air before he can transmit. In the real world, it's not as orderly as one client sends a packet and then the next gets to send with a zero delay. The net result is that when one client is sending data, the latency on the other clients increases because you are simply using up airtime.

This all differs from a fixed-frame TDM system where one client gets X amount of time and the second client gets Y amount of time and X+Y=100% because each client in such a system can never exceed the X or Y speed, respectively; but, the latency is more fixed.

As some other people pointed out, you can use bandwidth shaping (queues and traffic marking) to keep one client from using too much of the possible air time (read as possible throughput) thus keeping latency changes (jitter) to smaller values. You are limiting a client to throughput though so you may not like that solution if you thought you could sell 54Mbps to everyone.

There is no default configuration or a simple button to configure the systems as you need/like since we all use them differently. RouterOS is a blank canvas and you have to paint the picture you need.
 
User avatar
jwcn
Forum Guru
Forum Guru
Posts: 1501
Joined: Sun Aug 27, 2006 6:49 am
Location: Maryland, USA
Contact:

Re: POINT TO MULTIPOINT ISSUE

Mon May 12, 2008 11:59 pm

Excellent informative post. Thanks, I enjoyed it.
 
xbaha
Member Candidate
Member Candidate
Topic Author
Posts: 111
Joined: Fri Mar 09, 2007 12:16 pm

Re: POINT TO MULTIPOINT ISSUE

Tue May 13, 2008 12:00 am

Thanks for the extensive explination, it really help alot.
i still have few questions,

When doing BT between Client A CPE and AP CPE, as i stated, Client B pings become high, and that was explain by your previous reply.

i did the test again, and i notices, that also, CLIENT C & CLIENT D that are connected to another AP2 CPE, has higher pings, AP2 antenna is about 10 meters away from AP1, they all run on the same band, 2.4, but AP1 is on diffrent Channel than AP2...
is it logical with even diffrent channel AP2 and its clients gets effected?

also, i am thinking of adding another band on the 5Ghz, will it also be effected??

thanks alot..

tomorow ill try Nstream and see if giving up on latency enhance the performance.
 
User avatar
jwcn
Forum Guru
Forum Guru
Posts: 1501
Joined: Sun Aug 27, 2006 6:49 am
Location: Maryland, USA
Contact:

Re: POINT TO MULTIPOINT ISSUE

Tue May 13, 2008 12:04 am

Anytime you use btest it floods the radios. Increased latency will happen regardless of what client.
 
xbaha
Member Candidate
Member Candidate
Topic Author
Posts: 111
Joined: Fri Mar 09, 2007 12:16 pm

Re: POINT TO MULTIPOINT ISSUE

Tue May 13, 2008 12:44 am

regardless of channel and band???

so if i have 100 square meter floor tower, how much possible bandwidth i can get maximum there?

thanks.
 
User avatar
jwcn
Forum Guru
Forum Guru
Posts: 1501
Joined: Sun Aug 27, 2006 6:49 am
Location: Maryland, USA
Contact:

Re: POINT TO MULTIPOINT ISSUE

Tue May 13, 2008 1:09 am

I don't know that there is any way to answer that question. It depends on the frequency usage from other sources, signal levels and hardware you are using.

If there is one thing I can absolutely say about wireless it is there are no absolutes!
 
User avatar
Belyivulk
Member Candidate
Member Candidate
Posts: 285
Joined: Mon Mar 06, 2006 10:53 pm
Location: Whangarei, New Zealand
Contact:

Re: POINT TO MULTIPOINT ISSUE

Tue May 13, 2008 2:39 am

Agreed!

Wireless is more temperamental & variable than a woman’s mood!
 
User avatar
JJCinAZ
Member
Member
Posts: 473
Joined: Fri Oct 22, 2004 8:03 am
Location: Tucson, AZ
Contact:

Re: POINT TO MULTIPOINT ISSUE

Tue May 13, 2008 3:12 am

So yes, that should happen even with the 10 meters distance. Let me explain.

The 2.4GHz radio on the 802.11 card is sensitive to a range of frequencies from something like 2300Mhz to 2700Mhz (this is an example those aren't exact). The radio has a front-end tuner module on it which attempts to notch out the 20Mhz channel you are interested in. However, with the two antennas that close, they are probably in the near field to each other so when one AP is transmitting, the second AP's tuner module gets a big burst of energy into it and that overloads the tuner or makes it harder for it to do its job.

Think of it this way. If we were both standing 30 feet apart in an almost completely dark warehouse (we are the AP's in this example) and multiple other people (the stations) were at various distances through out the warehouse. Now to communicate we all have bright spotlights with different beam widths. The stations have 30 degree beam widths and us AP's have 90 degree beam widths. Now when you light up your spotlight to send a message to one or more of your stations, I (the other AP) am going to get hit in the eyes with stray light from you (even if I'm looking another direction). While that's happening, I have a harder time picking out the spotlights of my stations attempting to signal to me.

That's what's happening with your setup. There are a number of ways to solve it: cavity filters (but then you can't easily change channels), more spacing between AP antennas, precise synchronization of transmit between the AP's (not possible with 802.11), etc. Indeed, if your other AP was in an entirely different band (5GHz versus 2.4GHz) then you wouldn't have this problem.

So the simple answer to your question, is that when one AP is transmitting, the other AP will be affected by it and you should see it in increased latency and reduced modulation rates. Other people here may have better ideas on how to mitigate the problem, but it does make it hard on low latency apps like VoIP.
 
jonbrewer
Member Candidate
Member Candidate
Posts: 183
Joined: Sat Jun 05, 2004 5:56 am
Location: Wellington, New Zealand
Contact:

Re: POINT TO MULTIPOINT ISSUE

Tue May 13, 2008 1:08 pm

802.11 uses data frames and control frames. Control frames are transmitted using the Basic rate and the data frames are transmitted using the Supported Rates. This means that even if the link is "perfect" and able to do 54Mbps modulation at all times, you can't send 54Mbps through it because some of the frames are moved at 11Mbps. Further, there is frame overhead on the air, so you have further reductions in possible throughput.
Thanks for this really great post. I have never really understood "basic rate". It seems like the default configuration for Mikrotik is to have "basic rate" at 6mbps and supported rates a/g at 6-54mbps.

If we change "basic rate" to 54mbps only, will the control frames be transmitted at 64QAM (54mbps)?

Thanks,

Jon
 
User avatar
JJCinAZ
Member
Member
Posts: 473
Joined: Fri Oct 22, 2004 8:03 am
Location: Tucson, AZ
Contact:

Re: POINT TO MULTIPOINT ISSUE

Tue May 13, 2008 5:52 pm

Yes, you can customize the rates used and lock the units at specific modulations for Basic and Supported rates. This is how I run my PTP links -- locking at specific rates to get better latency and jitter. This doesn't work as well with PTMP setup's. On a PTP link, you (should) have highly directly antennas and stationary endpoints giving you a predictable path and set of conditions. On PTMP, the AP might have to talk with clients at a variety of conditions such as changing noise (the AP doesn't have a highly directly antenna in a PTMP setup so it hears more noise sources), varying distance to clients, varying client radios, different client antennas, etc. The AP needs to be able to use a lot of rates to be able to adjust down to the weakest client. You could lock the modulations for the LCD (lower common demoninator) but then all the clients perform at the lowest level.

Finally, you better have near perfect conditions to be able to lock at 54Mbps rate. I almost never see that work in real-world conditions.
 
xbaha
Member Candidate
Member Candidate
Topic Author
Posts: 111
Joined: Fri Mar 09, 2007 12:16 pm

Re: POINT TO MULTIPOINT ISSUE

Thu May 15, 2008 6:57 pm

i have used Nstream, pings went higher (30-50ms) but the bandwidth now is stable.
i am just worried about the maximum bandwidth that i can use in my place, doesnt seem to be that much, and probably about 25mbps per band to have stable connections... my plan is to have 200mbps !!

i was thinking of many ways to increase the total bandwidth in my hob... and i would like to ask:

can i lock the wireless card to listen to specific frequencies?
i.e. if the card can recieve from 2.1 to 2.5, and i requested a super channel @ 2.2, can i lock it to just listen to the 2.2 channels, and will not be effected by 2.1,2.3,2.4 etc.. ?

Other solution, if i used diffrent wireless card, say card 1 just listens to 2.2-2.3ghz and card 2 from 2.3-2.4, as i understood no card will be effected from the other card traffic, is this true?

Who is online

Users browsing this forum: No registered users and 27 guests