Many people involved in our craft understand when the marketing folks say something, it isn’t always that way in the real world.
One of those has to do with the ‘Gross‘ versus ‘Net‘ data throughput results with respect to Wireless LANs. Just because the stated data rate is 11Mbs, doesn’t mean you’ll actually get 11Mbs of net throughput.
For 802.11b – the max stated rate of 11Mbs works out around 5.5Mbs of net throughput. It is NOT the “half-duplex” nature of RF communications that is lowering the throughput, but the rest of the 802.11 protocol. Though many people make this incorrect assumption.
The ‘net’ throughput is brought down by overhead inherent in the protocol itself.
- Time wasted in the DIFS delays
- Time wasted in the Contention Window Delays
- Time and Bits wasted in Preambles
- Time and Bits wasted in SIFS/ACK delays
- Time and Bits wasted in RTS/CTS with their associated Preambles and SIFS delays
- Time and Bits wasted in large MAC Headers, etc.
- Time and Bits wasted in MAC Layer Fragmentation (Extra Headers)
Those are the real reasons you have a lowered net throughput.
In a typical packet exchange below lists those parts that are overhead – some are based on wasted time, and others on wasted bits – then there are parts that contribute to net throughput.
DIFS
Contention Window Delays
Preamble
RTS
SIFS
Preamble
CTS
SIFS
Preamble
MAC Header
Data frame payload
SIFS
Preamble
ACK
Only the actual payload of the Data frame contributes to the net throughput. Everything else in the packet exchange, time wasters like DIFS/SIFS and contention window delays, as well as all the other parts are just overhead. But it’s the use of the overhead that makes the 802.11 protocol actually work in a sometimes harsh RF environment.
It is a ‘ratio’ of those wasted time/bits to the total time/bits that takes you from gross bits in the air, to net bits of payload throughput.
This comes to bear with much more evidence as you delve into the 802.11n packet structures and their effects on throughput.
Again, it is NOT a result of only the half-duplex nature of RF communications – but the actual protocol itself.
For further information on this process, check out Marcus Burton’s great white paper 802.11 Client Arbitration.
I would appreciate your thoughts and comments on this subject. Thanks!
Thanks for the information Keith! I will have to refer people to this post when they ask why they are not seeing the throughput that they want!
Thanks for your kind comment. Glad to help. Anything else you'd like me to write a blog about? Perhaps some wireless question we can answer on the 'Wireless LAN Answers' upcoming podcast?
Thanks for your great post about wireless performance ! Are there any figures about 802.11 media utilization -vs- its impact on performance ? In a recent webcast, a vendor claimed that at around 35% media utilization, an AP performance will start decreasing. This seems low. If I recall the old days of HDX Ethernet CSMA/CD that threshold was around 60% and 80% for Token-Ring.
I don't have any details off the top of my head from reputable sources… but I agree that around 35%-40% utilization the retry rates seem to increase substantially. The performance decrease bit kicks in much earlier with CSMA/CA than with older Ethernet CSMA/CD.