Page 1 of 1

Help required for 20Mb 3 hop link

Posted: Thu Apr 14, 2005 1:11 pm
by RobClem
I need some help with which mikrotik hardware will perform best for a new customer link

I must first bring 40Mb over a ptp link with clear LOS.

I then must push at least 20Mb over a 10km ptp link with clear LOS to a repeater site

From the repeater site I must push the 20Mb over a hill to the customer who is 2Km from the repeater.

so in summary its 3 hops and I need a minimum of 20Mb available at the end of the link.

Can anyone tell me whats the best Mikrotik hardware for this link? i.e routerboard500/200 etc and which 5,8 radios to use.....



Posted: Thu Apr 14, 2005 4:49 pm
by mp3turbo2
you're going to have harsh times, dude. You may see some people post 50+ Mbps transfer rates, but those are bases on internal test-bandwidth test and probably not REAL tcp/ip transfer rates. My experience tells me that link which is able to show off 25Mbps in bandwidth test, has problems transferring real 7Mbit/s uncompressable data.

Posted: Thu Apr 14, 2005 5:29 pm
by stephenpatrick
Is that still true if you use EOIP to create a tunnel?
i.e. does re-encapsulating the data help by packing the data ?

I know EOIP uses CPU cycles, as does Nstreme, but use a faster CPU and you're there.

Also, I noted another user had throughput problems with "daisy chained" links, is that going to be a problem?

AFAIK on of the other users in Bulgaria has 50Mbps throughput using Nstreme at 30+km I think


Posted: Thu Apr 14, 2005 7:59 pm
by ghmorris
We lose a bit less than 50% on each hop. We get almost 30Mbs/30 miles with Nstreme Turbo on the first hop, no more than 19Mbs over 2 hops/60 miles etc etc.

This problem doesn't happen with Redlines, no one has been able to explain the per hop drop in bandwidth so far.


Posted: Thu Apr 14, 2005 9:06 pm
by RobClem
Stephen, we spoke about your new solution a couple of months back, have you tried them out in a real world environment?


Posted: Thu Apr 14, 2005 9:21 pm
by stephenpatrick
Hi Rob,

Yup I remember clearly.
Yes we have tested "real world" and done some extensive soak tests, but not at those sorts of distances.
We're scheduled to do some range-bandwidth tests soon, but believe it or not, for a wireless vendor, we've a lack of good sites (we're based in surburban south west London - doh!), and time on other big customer projects (laser) has delayed our radio range testing.

If you have a couple of towers/sites handy, and some big antennas for the long hop, we'll gladly loan you some equipment to test, including back-to-back as well as single link throughput.
That'll get things done super-quick, and tell you if this is going to be a winner or not ...

How does that sound?


Posted: Thu Apr 14, 2005 11:50 pm
by RobClem
sounds perfect to me. We are just finishing off testing another solution and if this does not do the job I will give you a call to test this link out.

drop me an email



Posted: Fri Apr 15, 2005 12:04 am
by stephenpatrick
great - there's one in your box ...

Posted: Fri Apr 15, 2005 7:41 am
by mp3turbo2
ghmorris, are your numbers real tcp/ip transfers or just bandwidth-test numbers?


Posted: Fri Apr 15, 2005 10:43 am
by djape
If you need 20mbit you should try you luck with RB500, ahteros 5211, at least 1.9m dishes...
And god help ya :)


Posted: Fri Apr 15, 2005 11:36 am
by mp3turbo2
you know, this should not be about luck. What are you going to do if this solution works right now and it will not be capable of delivering those 20Mbps in two months because of interference in the air?

High capacity business works other way.

Posted: Fri Apr 15, 2005 12:13 pm
by stephenpatrick
Well we are talking about unlicensed radio bands, so there's no guarantee of no interference ...
... but some of the problems people describe seem to be "system problems" certainly on back-to-back repeated links.
Those need to be tested live, and the known-good configurations posted so everyone can see it works


Posted: Sat Apr 16, 2005 4:36 am
by ghmorris
[quote="mp3turbo2"]ghmorris, are your numbers real tcp/ip transfers or just bandwidth-test numbers?


Real numbers.

There are caveats of course. MT does not like interference when delivering high speeds and there is no spectrum analyzer feature yet to track down problems.

The multi-hop speed degredation is a real problem.

We have also had some lockups running Nstreme on routerboards. We even had a lockup today on a P4 which surprised us.

MT really needs ATPC on a per-burst level for very long shots where both fast and slow fades are an issue. We're just getting into fade season here and will be shortly be replacing the MTs on our longest shots. They simply can't adapt to the changing RF conditions. Our experience is the Atheros cards don't like being shouted at or whispered to. They like a -55 to -65 signal for peak performance.

Saying all that, there is nothing else on the market we've found yet that packages this much throughput with this much control in an inexpensive package and we intend to continue with MT on short to mid-range high-speed links.