Forum

Hi All,

Can someone explain and discuss what is the acceptable throughput of 802.11g? I need a white paper also to prove to our customer to match with our results.

Please share everyone. Thank you very much.

802.11 b & g typically get about half the "head line" rate, ie the maximum rates, when operating under the best circumstances. So for b, that would be about 6 Mbps and 27 or so for g. Being a shared, half-duplex medium, that's all you should really hope for. Add in many client devices, or have too many beacons/second and you can cut those numbers down.

I have heard, but not tested it myself, that Apple products do slightly better, maybe 10 %.

I would agree with Wlanman's numbers, and possibly even suggest that 6 mbps (11b) and 27 Mbps (11g) are optimistic. 11g, with interference and a busy environment, expect less than half... maybe 20-24 Mbps.

This paper is all about mesh, but I was just reading it this morning (and had the link handy). They reference this 50% or less expectation and assume it in all the tables.

Top of page 7, http://www.strixsystems.com/products/datasheets/StrixWhitepaper_Multihop.pdf

Marcus,

I cheat !

I can run tests inside RF enclosures that keep out all the "noise".

But I agree, for normal installations I would consider my numbers as optimistic.

Dear All,

In my real testing per AP I can get only within a range of 14 to 20Mbps.

Thanks a lot for the input guys.

Are you limiting the AP to a particular rate, so that you can calculate the percentage ?

How is the interference level ?

Wlanman, im not sure the interference level. I forget to take it down. I have 45 AP's running on it using Aruba 3600 and AP105. What do i need to consider to achieve the best results? I'm sorry but I am new to wireless but 7 years experienced of wired. Thanks.

As it is half duplex this means a max theoretical send of 27Mbps and receive the same. But actual throughput will be lower than that due to interference, distance from AP and maybe other factors to do with the the power output and gain? So i think you'd get maybe 11Mb,18 or 24Mb? by the way would it be these exact speeds due to DRS?

Prash,

The first 27 is based on actual throughput. The latter numbers you mention 11, 18, and 24 are the so-called "headline rates". Yes, they are the actual rates, but we would expect about half those, or less, for the "real" throughput.

Dynamic Rate Selection will of course control the PHY rates, and by extension the throughput, but then you are getting into some things that are difficult to predict. That's one reason the conversation often deals with the [u]best[/u] expected throughput or rate. Client devices control DRS, and as Kevin has often said this is one of a manufacturers "secret sauces" - another being their Roaming algorithms.

Question, is that half-duplex throughput estimate of 1/2 the operating data rate considering other factors such as the use of protection mechanism, retries due to collisions and interferance, and fragmentation overhead? Throughput = 1/2 the data rate would seem to be best-case.