Maximum throughput on 802.11 links

Originally proposed in 2002 and officially published in 2009, the 802.11n standard has delivered more than double the throughput capability of previous 802.11a/g technologies by introducing the capability of multiple antennas.
 
Since (and even prior to) release of the official specification, a huge range of wireless capable equipment has become available, targeting a range of commercial consumer markets with various applications from indoor WLAN, outdoor point to point and point to multipoint.  Much of the technical documentation released for these products claim a maximum 'speed' of 300 megabits per second.
 
The first important point ot be mindful of is that this bit rate should never be confused with actual data transfer rates that you might expect to acheive, for example, when making a file transfer or watching a video stream across a wireless link.  This 300M bit rate often referred to is a measure of the transfer of bits between two devices on the network.  Due to protocol overheads, the actual data rate is quite different - and significantly less!  (for more info, refer to 'bit rate comparisons' link below)
 
Possibly a more useful measure of maximum throughput is actual network throughput, being a measure of the size of IP data packets carried across the link.  For 802.11n, we often quote a value of about 200 megabits/sec maximum throughput.
 
When we talk about 200mbps throughput note that we are referring quite explicitly to UDP traffic, and certainly only one direction (wireless, by definition, is not full duplex).  To measure this network throughput, we use data stream test tools such as iPerf (Wikipedia) or routerOS bandwidth Test. To achieve the maximum possible throughput, there are several requirements that must be met:
  • 802.11n protocol with dual channel operation (20MHz channel + 20MHz extension, look for ‘40MHz HT above/below’ in channel width)
  • Dual chain (MIMO) enabled (look on ‘HT’ tab, enable all 4 check-boxes)
  • At least 40dB signal to noise ratio (double click on link in wireless->registrations table, and review the connection statistics)
  • No other devices operating on the same channel and protocol (802.11 is designed to ‘share’ spectrum with other devices, so that it is possible to operate many networks with independent SSIDs using shared spectrum.  In that kind of environment, operation of other networks can impact on maximum throughput on others)
With all of these conditions observed, we have seen bench tests produce close to 220mbps throughput using MikroTik routerOS!
 
Even so, this is still not a throughput rate that you can expect to acheive in general use.  The first thing that you need to consider is that probably the significant majority of internet and network applications use TCP rather than UDP for data transfer.  TCP delivers more certainty of data transfer than UDP by making sure that receipt every packet transmitted between devices on a network are explicitly acknowledged by the recipient device.  That means that every packet of data transmitted results in a packet returned.  Since wireless packets can only be transmitted in one directiomn at a time, this requirement results in an approximate halving of the maximum throughput - thus maximum network throughput for TCP streams are about 100 megabits/sec.
 
Typically, realtime applications (voice, audio, video, etc) are UDP applications (though note that many of these types of services also have a TCP option) whereas other file based protocols (web, email, ftp, file/print service) are all TCP based applications.
 
So how does this network rate relate to actual useful payload throughput?  There are many reasons why you will rarely see that data speed when transferring a file, for example, across a wireless network:
  • When the network is point to multipoint, all client devices can try to transmit at the same time, causing interference, or collissions on the network, reducing throughput performance.  Proprietary systems (such as MikroTik nv2 and nstreme) use a special signalling technique that requires a station to receive 'permission' from the base station before transmitting.  Although this sort of mechnism helps to improve efficiency, transfer of those signalling messages does consume some of the total available bandwidth.
  • Application overhead: often, applications pass 'out of band' data across the network, such as file names, bytes remaining, data integrity information and so forth.  This extra information can increase the total amount of data that is transferred across the network, and 
  • Frame packing: data is transmitted in set chunks, and not all packets are necessarily full!  The net result is that payload is less than optimum.
  • Wireless is prone to data loss, due to interference, reflections, atmospheric dispertion and fluctuations in density etc, and so also impacts on the total actual payload throughput.
In the real world, we have generally observed actual TCP payload throughput across wireless p2p links to acheive up to about 70 megabits/sec under ideal conditions.  For longer distance links (>2Km) in relatively urban areas, the throughput expectation is nearer to 25mbps TCP.
 
These are the kinds of data rates that you can reasonably expect to acheive in real world applications.
 
If you are interested in some further reading, the following links may be useful: