Community discussions

MikroTik App
 
ejansson
Member
Member
Topic Author
Posts: 301
Joined: Fri Oct 21, 2005 4:09 pm
Location: Manitoba, Canada

Routing AP's and Nstreem Polling

Fri Nov 03, 2006 6:14 pm

Who out there is using routing to the wireless card rather then bridging/wds ?
I'm guessing there should be a nice bump in performance due to reduced overhead. Everyone seems to be just using bridging. Any reason or is it just that it is simpler network.

Part 2

I'm also interested in people experience with using Nstreem Polling with an ap and clients. Again I am assuming there should be a performance gain due to the polling as it reduces channel contention as the collisions associated with hidden nodes are eliminated

Thanks in advance for your replys.

Erik
 
galaxynet
Long time Member
Long time Member
Posts: 648
Joined: Fri Dec 17, 2004 2:52 pm
Contact:

Sun Nov 05, 2006 3:50 pm

We use Nstreme and routing exclusively on all our high bandwidth APs/clients (business class).

We use straight routing in all our 'consumer' (vice business class) AP cells.

Depending on what you are using for APs (non-mikrotik MB or Mikrotik MB) you could run up against max CPU utilization at 20 - 30Mbps (Mikrotik 5XX MB) with Nstreme protocol. We have seen in excess of 50Mbps with non-Mikrotik MB running 600mhz processors (and higher 1.7ghz PentiumM) using Nstreme and routing.

We have one dual or Nstreme2 setup that can consistantly handle 60+mbps - haven't seen it go any higher - client loading not lab loading. We use two differnt channels w/two different antennas (5.2 and 5.8ghz channels) w/vertical and horizontal polarizations seperated by about 6 feet.

We use to use bridging modes but you are right there is some gain in performance w/using routing vice bridge, however you can lose some due to just raw cpu processor power (speed) especially w/using Nstreme protocol.

Best advice is to try both - for us it works better and makes a cleaner network w/we use routing - w/or without Nstreme.
Thom
 
karyal
Member Candidate
Member Candidate
Posts: 168
Joined: Sat Nov 12, 2005 12:09 pm

Sun Nov 05, 2006 7:14 pm

Depending on what you are using for APs (non-mikrotik MB or Mikrotik MB) you could run up against max CPU utilization at 20 - 30Mbps (Mikrotik 5XX MB) with Nstreme protocol. We have seen in excess of 50Mbps with non-Mikrotik MB running 600mhz processors (and higher 1.7ghz PentiumM) using Nstreme and routing.
uhm.. what's the clients distance/signal? we have several AP (around 60) in place, no nstream, no turbo we can't go higher than 10MBps (hdx), and no more than 20 clients running to keep pings low (less than 10 ms)
This is with a single routerboard running two cards, one or the PtP link and one for PtMP links.
We don't need more bandwith at this moment, but i'm really curious...
Bye,
Ricky
 
galaxynet
Long time Member
Long time Member
Posts: 648
Joined: Fri Dec 17, 2004 2:52 pm
Contact:

Mon Nov 06, 2006 5:15 pm

I take from the above posts though that you are using bridging not routing - correct?

Using WDS eliminates some of the overhead (CPU overhead) that you would see if you used Nstreme - but you lose some bandwidth - check the forum on that - I believe that you lose about half your avaialble bandwidth. And if you're using bridging then subtract 10 - 30% for bridge overhead traffic.


Clients / distance - where we have the 60Mbps backhaul...it's a link that is shot over about 8 miles to a business park, from there we use two 800Mhz cpus w/MT OS to service the 14 business clients that are all bandwidth 'hogs'. The 60Mbps link is Nstreme2 w/two cards and antennas. In the park we use Nstreme and routing here - especially since we are handling public IPs for each and every client and they all have some kind of server running (mail/web/etc). Distance in the park is < 1/2 mile. Ping times can vary a litlle w/loading but they are consistantly very low over the entire link setup. Especially w/testing only the Nstreme2 link (it's full duplex so it's FAST no matter how loaded it is).

We use a very fast (relative now) CPU - 1.8 Pentium M processor (cooling and power consumption are a major considerations for these 'heavy duty' cpu setups).

For our 'consumer' setups we keep the pings <20ms. This is because we have some pretty long links out there...the longest is; 2 miles from homeplate to first base, 10 mile backhaul to second base, 14 miles to consumer. When you look at distances like that you incur distance penalties just because it takes time to actually travel the distance plus each 'base' adds it's own latency.... At the AP cells themselves we look at as short a ping time as possible but from 'homebase' it's relative to distance and number of 'jump points' in between....

Signal levels...run from in the high -40s to -80db, past -80 we either look at adding a cell in the area or don't service the client - better not to service them at all than give them bad service.

We use alot of MT 5XX series routerboards at the APs, they're inexpensive but very flexible and robust units. When they get loaded we split the cell and add another MT 5XX routerboard sector. After that we go to the 'bigger' cpu units. We only have one (consumer) cell like that though the rest are split up using MT 5XX boards.

Hope all this is heplful to you.

Who is online

Users browsing this forum: No registered users and 43 guests