Community discussions

MikroTik App
 
changeip
Forum Guru
Forum Guru
Topic Author
Posts: 3830
Joined: Fri May 28, 2004 5:22 pm

Antenna and distance to clients

Sat May 14, 2005 11:30 pm

I am toying with the idea of providing internet access from atop our mountain here in southern california. I have the possibility of reaching 100,000 homes at our location. In the discovery I am wondering what the broadcast distance from AP to client would be for the following types of antennas:

Horizontally Polarized 95 Deg Sector Antennas
http://www.pacwireless.com/products/pawsa24.pdf

or Omni's
http://www.mikrotik.com/Documentation/O ... tennas.pdf

I am curious about the omni ranges - can you get a 5-10 miles out of them with a directional client antenna or is that dreaming? All clients would really be within 120-180 degrees of the AP.

If we use a sectoral antenna will clients within a few miles of each other be able to still connect to the AP, or would it be limited to a smaller footprint on the ground below?

Here is a link to a picture of the area:
http://pictures.changeip.com/NorrisFire ... rea?full=1

Here is a link to the google map showing this area as well:
http://maps.google.com/maps?q=32.966366 ... &t=k&hl=en

Just wondering if I will need more than 2-3 antennas on the mountain top to reach the area outlined on the picture above. And is providing wireless to clients 5-10 miles away worthwhile? Line of sight is visible from almost every home on that map : )

Thanks all
Sam

Mikrotik rules!
 
UniKyrn
Member Candidate
Member Candidate
Posts: 245
Joined: Fri Dec 24, 2004 9:27 pm
Location: Spokane, WA

Sun May 15, 2005 3:03 am

If everybody is in that 120-180 degree arc, I'd put a good waveguide antenna up there. I've seen problems with omni's up high like that, while they can see for miles, the problem is "they can see for miles", they get to hear all the noise from every source imaginable. Add to the fact they are lower gain then the waveguide and you're wasting half your power in a direction that is useless, and the extra cost for a good antenna might be money well spent.

Once you get up there though, clients 5-10 miles away shouldn't be a problem.
 
randyloveless
Member Candidate
Member Candidate
Posts: 207
Joined: Thu Sep 30, 2004 10:14 am
Location: california
Contact:

Sun May 15, 2005 3:37 am

the other thing you have to look at is amount of cpe unit you will have .

lets just say you want 500 clients . there is no way you can put all of them on 2 or 3 radios .

to much load

Randy
 
changeip
Forum Guru
Forum Guru
Topic Author
Posts: 3830
Joined: Fri May 28, 2004 5:22 pm

Sun May 15, 2005 3:46 am

So is it safe to say that you cannot provide wireless internet to 500 clients on a single machine - no matter the processor size and number of wireless cards? Are there not many people providing wireless internet to residential customers?

Sam
 
randyloveless
Member Candidate
Member Candidate
Posts: 207
Joined: Thu Sep 30, 2004 10:14 am
Location: california
Contact:

Sun May 15, 2005 7:52 am

i am doing this but not from 1 location . we sectors and spilt a area up . you could most likely get maybe 50-100 units per radio .

the main reason you cannot have so many people or clients on one radio

is thru put lets say you have 50 cpe's going to one radio and lets say your
customers are all with in 3miles you are giving them what is called best effort service . which basicly means you will give them oh 1.5mb thruput
and if half of them are online at once your radio thruput in going to suck.

its a gamble but it would be safe to say 50 heavy users would be about the limit of each radio.

the faster the processor the better performance you clients will get . but i still think if you go over 100 cpe's per radio is too much..

also how much bandwidth are you going to have at you main location

Randy
 
UniKyrn
Member Candidate
Member Candidate
Posts: 245
Joined: Fri Dec 24, 2004 9:27 pm
Location: Spokane, WA

Sun May 15, 2005 10:26 pm

There is another limitation that come in for 802.11b, it's a polling protocol and the more radios that are registered, the more polling that gets done. Past a certain point the radio is so busy polling that it can't pass traffic efficiently and still poll everybody.

As a general rule, I don't put more than 64 customers on any given radio. That might be conservative, but it's worked fairly well.
 
randyloveless
Member Candidate
Member Candidate
Posts: 207
Joined: Thu Sep 30, 2004 10:14 am
Location: california
Contact:

Mon May 16, 2005 12:38 am

thats what we use to do with all are karlnet radios or should i say they had a limit of 64 client connection . reggie was working on before they were bought out on a 128 client version dont know what ever happened to it we tested it but it was not stable past 90 clients this was with 2.4 silver card . dont know if the newer cm9's would handle it better the lucent cards

Randy
 
GJS
Member
Member
Posts: 418
Joined: Sat May 29, 2004 4:07 pm
Location: London

Tue May 17, 2005 1:17 am

Just to clarify, 802.11b is a *collision* based MAC, polling is what you actually want in preference.

Google for "hidden node" and you should find plenty on this.

OT: Is it possible to tell when you are getting close to the maximum number of clients on an AP without waiting 'till you are over it and performance takes a dive?
 
UniKyrn
Member Candidate
Member Candidate
Posts: 245
Joined: Fri Dec 24, 2004 9:27 pm
Location: Spokane, WA

Tue May 17, 2005 1:26 am

Hummm, I thought I remembered reading it was a polling protocol, sorry.
 
User avatar
stephenpatrick
Forum Veteran
Forum Veteran
Posts: 702
Joined: Fri Aug 20, 2004 12:26 pm
Location: UK
Contact:

Tue May 17, 2005 10:11 am

If you only have the one location, it **should** be possible to put multiple radios in a box (or for that matter, different boxes) and put them on separate (non-adjacent or overlapping) frequencies, pointing separate antennas at the same coverage sectors.
Doesn't help with "overall coverage" which might be better with multiple base sites, but you might be constrained otherwise. Obviously there are a lot more frequencies at 5GHz to avoid self-interference doing this.
I'm guessing you use separate SSIDs on each radio to regulate the loading per radio, but other users might want to comment here - is there another way?
Totally agree, sectored base sites are best, with traffic loading as well as interference and range considerations.
What do people prefer most - 90 or 120degree sectors?

Regards
 
User avatar
djape
Member
Member
Posts: 465
Joined: Sat Nov 06, 2004 7:54 pm
Location: Serbia

Wed May 25, 2005 9:17 pm

Think about pole with 5 sector 802.11a antennas and Pentium IV with max GHZ and Ram, and you can reach arround 350 users :)
BUT!!! If I could have chance to do service for more than 500 users I would do 5 Mikrotiks without thinking!
Use 802.11a and sector on different frequencies and I can bet you can go 100+ per AP with atheros.

Currently i reached ~70 users per AP on 802.11b without expiriencing slownes, but no p2p and communication between users is off...
Block p2p, online gaming, VOIP, limit max connections and only god knows how many users you can serve.
Also it is very important what kind of service you want to provide???
If you want to give them high quality of service, forget for whole thing....
If you want to give them browsing, download, mail, chat, it's cool, you can do it :)

Cheers and let me know what happened...
 
randyloveless
Member Candidate
Member Candidate
Posts: 207
Joined: Thu Sep 30, 2004 10:14 am
Location: california
Contact:

Thu May 26, 2005 10:30 am

hey all

on this subject . we need to replace all our karlnet equipment and want to use MT but we have heavy users on system mostly business clients. 2mb plus per user about 20 user per site . karlnet ap1000 work well most of the time. be testing MT for about 4 months and worried about performance and qos . now with MT we can filter and give better qos .

any suggestions . looking at using 1 ap per 20-30 heavy users with p4 3ghz system working on kooling issue in system

Randy
 
User avatar
stephenpatrick
Forum Veteran
Forum Veteran
Posts: 702
Joined: Fri Aug 20, 2004 12:26 pm
Location: UK
Contact:

Thu May 26, 2005 11:14 am

Pentium M 2GHz, much cooler than P4.
Passive/active cooling is possible, if you've enough time/patience and prepared to spend the NRE on R&D.

I heard that AMD "mobile" CPUs will fit normal AMD motherboards.
Is that the case?
If so, total solution probably cheaper than Pentium M.

Anyone got any experience ?

Regards
 
User avatar
djape
Member
Member
Posts: 465
Joined: Sat Nov 06, 2004 7:54 pm
Location: Serbia

Thu May 26, 2005 5:48 pm

Well, if you want to save few dollars try, but personally I wouldn't go for it...Intel is more stable...
 
randyloveless
Member Candidate
Member Candidate
Posts: 207
Joined: Thu Sep 30, 2004 10:14 am
Location: california
Contact:

Thu May 26, 2005 7:40 pm

amd processors have a greater heat output then intel . plus they can have issues of locking up for no reason other then they just wanted to cause you a headache . i had tested some amd processors they do work but are way hotter then a intel so if your using them on a tower forget it . in a controled room maybe . i would still go with intel.


Randy
 
randyloveless
Member Candidate
Member Candidate
Posts: 207
Joined: Thu Sep 30, 2004 10:14 am
Location: california
Contact:

Thu May 26, 2005 8:22 pm

what motherboards have you all been using for the mobile processors
 
sten
Forum Veteran
Forum Veteran
Posts: 919
Joined: Tue Jun 01, 2004 12:10 pm

Fri May 27, 2005 1:47 pm

Well, if you want to save few dollars try, but personally I wouldn't go for it...Intel is more stable...
Intel is more stable? With Mikrotik routeros or is it just a more stable platform?
 
User avatar
djape
Member
Member
Posts: 465
Joined: Sat Nov 06, 2004 7:54 pm
Location: Serbia

Fri May 27, 2005 4:42 pm

With Mikrotik Router OS. I'm talking about strange 100% processor usage etc... My personal PC is AMD, but it's just my PC :) For something serious I wouldn't use AMD ;)
 
User avatar
stephenpatrick
Forum Veteran
Forum Veteran
Posts: 702
Joined: Fri Aug 20, 2004 12:26 pm
Location: UK
Contact:

Fri May 27, 2005 9:20 pm

FWIW,
I have had lock-ups once on AMD, many times (and consistently) on VIA with standard BIOS, and none on Pentium M and on custom VIA BIOS boards.

Don't think that's definitive in any way, but getting a stable CPU/mobo/MT combination isn't always simple.
I'd blame the mobo/Bios in most cases.

Regards
 
v1nt3r
just joined
Posts: 6
Joined: Tue Feb 01, 2005 11:31 am
Location: Jozefow, near Warsaw, Poland
Contact:

Tue May 31, 2005 3:47 am

We're usign AMD Semprons 2200+ and 2300+ on ASUS A7V8X-X (with 128 MB RAM) (4 base stations, 1-4 atheros'es + 0-1 prism's 2.5 on each). No problem (only this heat now... summer is comming!). Be sure to take care about cooling (today - 36 celsius in shadows, inside of MT boxes on stations - ~50 celsius). I've noticed problems with other MB: AsRock KT-something. Those two MB's are based on KT400, but AsRock wasnt stable - so we've changed it (its simple - ASUS is STABLE).

Clients make 4-5 mbit/s on each - we've got some rules on firewall - cpu load floats about 1-9 percent. On AMD Celeron 600 - about 25-60 percent (... and sometimes higher pings). Tested on MT 2.8.13, 2.8.27, 2.9rc1, rc2 and now working on MT 2.9rc4.

I recommend Athlon Mobile 2000+. It can work having 100 celsius temp.! Semprons arent so hard - they'll work only in 0-60 celsius temp.

Sorry for my english ;-)

Greets for everybody fighting with MT boxes!
 
jonbrewer
Member Candidate
Member Candidate
Posts: 182
Joined: Sat Jun 05, 2004 5:56 am
Location: Wellington, New Zealand
Contact:

Wed Jun 15, 2005 8:31 am

We're usign AMD Semprons 2200+ and 2300+ on ASUS A7V8X-X (with 128 MB RAM) (4 base stations, 1-4 atheros'es + 0-1 prism's 2.5 on each).
Czesc v1nt3r! We use some A7V8X with Sempron down here in Nowa Zelandia. They have all been running great since install. Just as good as our Intel boards at half the cost. :-)
 
variable
Member Candidate
Member Candidate
Posts: 217
Joined: Wed Apr 13, 2005 4:36 am

Wed Jun 15, 2005 9:46 am

amd less stable -bah!; in my experience amd is much more stable, all my servers, linux+mt run amd. never had a crash. the people i know using intel have lockups always! and as far as heat. AMD is much much much cooler, using similar heatsinks and frequencies, amd runs around 20degrees farenheit cooler.

and yes it is true that mobile amd procs(socket a) work in desktop boards, which means a much cheaper system as well!


i currently use mostly durons and semprons
 
randyloveless
Member Candidate
Member Candidate
Posts: 207
Joined: Thu Sep 30, 2004 10:14 am
Location: california
Contact:

Wed Jun 15, 2005 9:51 am

any body know why my mobile amd 2400 read on the motherboard it is a 600? am i setting it up wrong

Randy
 
wildbill442
Forum Guru
Forum Guru
Posts: 1055
Joined: Wed Dec 08, 2004 7:29 am
Location: Sacramento, CA

Wed Jun 15, 2005 9:34 pm

Like djape suggested efficently managing your bandwidth will allow for greater number of users per AP.

What it all comes down to is you're trying to service 100+ users off of a ~5mbps connection (802.11b) so bandwidth shaping is needed otherwise your power users will bring your network to a halt. Also as someone also mentioned 802.11 is a CSMA/CD protocol so collisionsn and retransmissions are going to happen cutting into that 5mbps available to users. I setup configured data rates on all our AP's so that clients cant associate with higher than a 2mbps data rate since that is our SLA.

On my networks I do a combination of protocol based QoS bandwidth limiting as well as setting MIR's for each customers level of service. Limiting P2P Uploads to 128kbps, capping total P2P downloads to 2mbps, and giving VoIP and HTTP traffic priority to the pipe has greatly increased the performance on our network. I don't see why anyone would want to limit users from using VoIP, we're trying to bundle VoIP with our internet service and encourage users to get away from the traditional PSTN based services.

IP traffic is bursty in nature, packet sizes vary, and theres no linear relationship between the number of users and available bandwidth.
 
variable
Member Candidate
Member Candidate
Posts: 217
Joined: Wed Apr 13, 2005 4:36 am

Wed Jun 15, 2005 10:43 pm

well as one of my users pointed out. who will i call if my internet goes down?
 
v1nt3r
just joined
Posts: 6
Joined: Tue Feb 01, 2005 11:31 am
Location: Jozefow, near Warsaw, Poland
Contact:

Thu Jun 16, 2005 5:01 am

Our discussion has got OT status ;-)

In our network we are building bridge's using IEEE 802a and g (D-Link DWL-G520 PCI and unknown atheroses on 5213 - miniPCI, ~18dBm in 5GHz) with IRIS-1 (parabolic antennas, 19dBi, short RG58ALL or H1000 cables, 2312-2402 MHz) or Genetrix Parabolic (parabolic antennas, 21-24dBi) in b or GigaEther (directional antennas, 23dBi) with Genetrix Sectors (in future, now we've got directional-to-directional links).

In "g" links are from 1,1 km to 2,25 km. Links varies from stable 36 mbits (2250 meters, about <20% fresnel) to 54 mbits (two shorter ones with >80% fresnel). Links in "a" are from 2,5 km to 5,7 km. Links are stable 54 mbits. Real througput with "g" bridge is about 4-8 mbits full-duplex, with "a" bridge is about 16+ mbits full-duplex.

Clients are connecting to sectors and omni's (links from 50 meters to 4,5 km) connected to D-Link DWL-G520 (mixed mode b/g) or ZCOMs prism's 2.5 200mW (only b). Real througput is about 3-4 mbits full-duplex on each interface. Sometimes there isnt any LOS ;-)
I am curious about the omni ranges - can you get a 5-10 miles out of them with a directional client antenna or is that dreaming? All clients would really be within 120-180 degrees of the AP.

If we use a sectoral antenna will clients within a few miles of each other be able to still connect to the AP, or would it be limited to a smaller footprint on the ground below?
To changeip:
If you want to build big wifi network i recommend:
1. IEEE 802.11a, links in 5GHz, Mikrotik vs AP (for example Repotec has got cheap AP with antenna connectors);
2. One sector antenna (19-21dbi) with for example CM9 for real 10-20mbits of full-duplex traffic, or for ~50-100 clients;
3. Maximum of 3 sector antennas/"a" interfaces for one Mikrotik base station;
4. ALWAYS LOS when using 5GHz :D
5. BS based on standard ATX boxes, i prefer Semprons, you should use something powerfull enough to take care those bandwiths ;-) (minimum 1500MHz i think);
6. Routing and firewall rules on no-wifi Mikrotik box (connected with standard ethernet to base stations).

In this case you will get about 30-60mbits with 150-300 clients connected to one BS.

...but this is only a theory :P. Real world always seems to redefine your plans!

Sorry for my english...
 
v1nt3r
just joined
Posts: 6
Joined: Tue Feb 01, 2005 11:31 am
Location: Jozefow, near Warsaw, Poland
Contact:

Thu Jun 16, 2005 5:57 am

Our discussion has got OT status ;-)

In our network we are building bridge's using IEEE 802a and g (D-Link DWL-G520 PCI and unknown atheroses on 5213 - miniPCI, ~18dBm in 5GHz) with IRIS-1 (parabolic antennas, 19dBi, short RG58ALL or H1000 cables, 2312-2402 MHz) or Genetrix Parabolic (parabolic antennas, 21-24dBi) in b or GigaEther (directional antennas, 23dBi) with Genetrix Sectors (in future, now we've got directional-to-directional links).

In "g" links are from 1,1 km to 2,25 km. Links varies from stable 36 mbits (2250 meters, about <20% fresnel) to 54 mbits (two shorter ones with >80% fresnel). Links in "a" are from 2,5 km to 5,7 km. Links are stable 54 mbits. Real througput with "g" bridge is about 4-8 mbits full-duplex, with "a" bridge is about 16+ mbits full-duplex.

Clients are connecting to sectors and omni's (links from 50 meters to 4,5 km) connected to D-Link DWL-G520 (mixed mode b/g) or ZCOMs prism's 2.5 200mW (only b). Real througput is about 3-4 mbits full-duplex on each interface. Sometimes there isnt any LOS ;-)
I am curious about the omni ranges - can you get a 5-10 miles out of them with a directional client antenna or is that dreaming? All clients would really be within 120-180 degrees of the AP.

If we use a sectoral antenna will clients within a few miles of each other be able to still connect to the AP, or would it be limited to a smaller footprint on the ground below?
To changeip:
If you want to build big wifi network i recommend:
1. IEEE 802.11a, links in 5GHz, Mikrotik vs AP (for example Repotec has got cheap AP with antenna connectors);
2. One sector antenna (19-21dbi) with for example CM9 for real 10-20mbits of full-duplex traffic, or for ~50-100 clients;
3. Maximum of 3 sector antennas/"a" interfaces for one Mikrotik base station;
4. ALWAYS LOS when using 5GHz :D
5. BS based on standard ATX boxes, i prefer Semprons, you should use something powerfull enough to take care those bandwiths ;-) (minimum 1500MHz i think);
6. Routing and firewall rules on no-wifi Mikrotik box (connected with standard ethernet to base stations).

In this case you will get about 30-60mbits with 150-300 clients connected to one BS. You should get about 10+ km links when using 19-21dBi sector antennas and 23+ dBi directional client antennas.

...but this is only a theory :P. Real world always seems to redefine your plans!

Sorry for my english...

Oh, i forgot. In out network clients are using IRIS-1 (parabolic antennas, 19dBi) for long range links, ATK16 (http://dipol.com.pl/sklep/a712427.htm) for medium range links and small panels (8dBi) for short range links. All clients are connected using Planex GW11SP IEEE 802.11b Access Points (1 RPSMA connector, 2 eth ports, based on OvisLink AP).
 
mp3turbo2
Member Candidate
Member Candidate
Posts: 196
Joined: Wed Jun 02, 2004 9:15 am

p2p limiting

Thu Jun 16, 2005 11:28 am

> Limiting P2P Uploads to 128kbps, capping total P2P downloads to 2mbps

... etc, etc.

wildbill, how did you do that? I created thread for p2p capping, but nobody was willing to share his knowledge with me. I am able to disable p2p totally, I'm able to limit p2p download and upload to the same value, but I was not able to differentiate upload from download without distortion of p2p service (download was behaving like PCQ but it should not when I limited _upload_).

Will you be so kind and share "those" six lines of your config? :)

thnx, mp3turbo.
 
randyloveless
Member Candidate
Member Candidate
Posts: 207
Joined: Thu Sep 30, 2004 10:14 am
Location: california
Contact:

Thu Jun 16, 2005 12:09 pm

p2p up and down should just create a mangle rule p2p-up out wan
and p2p-down in lan

i will have to look at our setup but that should help .you may want to ask around for better answer

Randy
 
wildbill442
Forum Guru
Forum Guru
Posts: 1055
Joined: Wed Dec 08, 2004 7:29 am
Location: Sacramento, CA

Thu Jun 16, 2005 8:12 pm

yeah basically what randy said.. its in the mangle rules, when I get back to my home office I'll copy the config over to show you..
 
mp3turbo2
Member Candidate
Member Candidate
Posts: 196
Joined: Wed Jun 02, 2004 9:15 am

Fri Jun 17, 2005 10:38 am

if you can, that would be excellent. Thank you.
 
wildbill442
Forum Guru
Forum Guru
Posts: 1055
Joined: Wed Dec 08, 2004 7:29 am
Location: Sacramento, CA

Fri Jun 17, 2005 9:04 pm

13   ;;; Peer 2 Peer - IN
     dst-address=66.60.xxx.xxx/xx p2p=all-p2p action=accept mark-flow=p2p-in 

14   ;;; Peer 2 Peer - OUT
     src-address=66.60.xxx.xxx/xx p2p=all-p2p action=accept mark-flow=p2p-out 
You could also do this by specifying the interfaces...
6   ;;; P2P-IN
     in-interface=WAN p2p=all-p2p action=passthrough mark-flow=p2p-in  

 7   ;;; P2P-OUT
     in-interface=LAN p2p=all-p2p action=passthrough mark-flow=p2p-out 
its nothin fancy..

Who is online

Users browsing this forum: anav and 46 guests