5054 shows high rate but slow throughput

[Velocity]

I have a 5054 link with Echo 17db directionals .5miles apart clear LOS.
The highest I have seen on the throughout test is 1500KB. the rate shows 108 or 54 in turbo or non-turbo respectively.
The signal has gone from -69 to -77
I have tried changinf ack from 100 to 400
Tried changing TX power from 13 to 20

I hope to have some 5354 cards soon just in case it is a compatibility issue.

Any Ideas?

Thanks,
Rodger

[dkii]

I'm not sure what the recieve sensitivity is for 108Mb, but -68 is the minimum for 54Mb.. I definately wouldn't expect -77 to pass 54Mb. Also, what type of system are the cards in? I think they require about 500mhz to achieve full throughput.

[lonnie]

The throughput is not limited by the cards. The limiting factor is the CPU. It takes a 500 PIII or better to max out the data.

[mp3turbo]

dkii is also right, maybe after manually locking 36mbps, 24mbps or even 18mbps you'll get higher numbers, velocity... also, antenna aiming definitely is very important factor - you can get easily confused by having sidelobe and considering it to be the main signal.

simply said, there are too many variables. Let's don't mention pigtails, antenna cables, their quality...

bye, mp3turbo.

[Velocity]

It is in a WRAP Board. Has anyone done any throughput testing with these boards.

[tony]

The WRAP boards (Geode 233-based) have a throughput of roughly 3,000KB/sec routing (via ethernet or Atheros).


If you are running a throughput test from a WRAP, or to a WRAP, the results will be lower due to heavy CPU utilization during the tests.

[lonnie]

The thing we do is lock to 6 mbps and we still see our 2-3 MByte/sec. If you lock too high it does not scale down, but locking low allows it to scale up very nicely if the signal is there.

[szern]

The WRAP boards (Geode 233-based) have a throughput of roughly 3,000KB/sec routing (via ethernet or Atheros).
That would be the equivalent of 24Mbps?
The WRAP boards Valemount ships are 233MHz or 266MHz?
Does it make a big difference? (the MHz rating)

How would it (WRAP board) fair if I'm using both Mini-PCI slots (with Senao b/g cards) operating with Radio 1: Client mode (linking to my Wireless Backhaul) and Radio 2 : AP Mode (serving local users)
Effectively a Microcell POP configuration. What sort of throughput could I expect?

Thanks!

[lonnie]

So far we have only used the WRAP board for dual backhaul systems. Suffice to say that we can easily fill a 4 mbps ADSL pipe. We have nothing faster than at this time so two radios, each capable of 20 mbps, are sharing a 4 mbps line. No problem from what we see.

I have not yet put any repeater systems into place so I cannot answer this yet. Put it this way - it will provide more throughput for less $$ than a lot of the traditional units out there. It is a good place to start.

Most of you guys are coming from less than 5 mbps services and the 5.x GHz is mainly to give you more channels. If you need 30 mbps full duplex, then you have to talk to Proxim or Redline.

Our boards are 233 MHz processors as Tony said Geode 233-based.

[szern]

We're looking to deploy the WRAPs in a Private Wireless Network type of situation (Internet Access is there, but its use is secondary), the primary use is for dispersing bandwidth-intensive multimedia files to a number of locations.

A typical scenario would be like this (a mixture of both):

1) [Server]---[Client Radio]---[Radio1-MAIN BASE STATION-Radio2]---[Radio1-MicroCell-Radio2]---[Client Radio]---[Client PC] <--- MicroCell functions like a Repeater (Relay Point) - Relaying signal for a Client PC with NLOS to Main Base Station
2) [Server]---[Client Radio]---[Radio1-MAIN BASE STATION-Radio2]---[Client Radio]---[Client PC] <--- Direct Connection (For Client Radios with LOS to Main Base Station)

* Radio 1 & 2 are Senao 802.11g Mini-PCI Cards (802.11g Mode, H-Pol)
* All Boards used are WRAP (Base Station, MicroCell POP, Client Radio)
* On average, each Base Station would have 8-9 Direction Connections/MicroCell connections
* There's only 2 MicroCell (Repeater/Relay Point) with a maximum of 2 Client PC's each
* Typically, each Client PC, requires about 3-5Mbps bandwidth

Would having 2 Radios on each WRAP (Base Station & MicroCell POP) degrade performance?
In effect becoming sort of Half-Duplex (if you can call it that) due to store-and-forward type of situation?

Thanks!

[lonnie]

As long as the radios are Atheros based we see no major decrease in the store and forward mechanism, except for slightly higher latency. The driver is DMA based and does not take CPU resources, so two radios can be doing simulateneous use.

The prism2.x series are polling and therefore you will have the system tied up when one radio is in use. For that reason you see a major loss of throughput when you have more than one radio and one of them is a prism2.x

[dkii]

prism 2.x polls?!?!?!?

[tony]

No, just it's method of accessing the card. (PIO vs DMA)

[dkii]

Ahh I see.. i/o polling, not AP polling..