The Norwegian research network
  
  Search:

Throughput

Big numbers sell better. This is also the case for wireless networks. A "11Mbps" or a "54Mbps" wireless network might sound impressive but it is not what you as a user will experience. A lot of the capacity is lost to communications overhead and the data rate given is what can be expected of the physical layer. Wired networks also give their capacity on the physical layer but the overhead is not that large and the experienced throughput is close to the physical capacity.

To understand where the overhead is, it is a good idea to understand about the network layers. This table shows the OSI reference model and how it relates to the world of TCP/IP. The light yellow boxes show the wireless specific layers.

 

When transmitting a packet of data which comes from the application layer, headers has to be added for each layer, making the total transfer size larger.

The following shows how the IEEE 802.11b operates:

This is an example of how headers are added. On top of the data comes the TCP header of 20 bytes (assuming the option fields are not used). Then comes the IP header of 20 bytes. Together they make the TCP/IP datagram. The size of this the same as the MTU, usually 1500 bytes to get through most router configurations. Then comes the SNAP header (used to encapsulate original Ethernet V2 ethertype values into an IEEE 802.2 frame) of 5 byte and the LLC (Logical Link Control) header of 3 bytes. This is in turn encapsulated by a MAC header of 30 byte and a trailing FCS of 4 byte. This is the MPDU or the PSDU part of the PPDU. Then we get a PLCP header of 6 byte and finally a PLCP preamble of 18 byte ("long preamble"). The preamble is used to signal that "here is a train of data coming" to the receiver. The 802.11b standard gives an option of reducing the size of the PLCP preamble to 9 bytes ("short preamble"), significantly increasing performance on the higher rates.

This is what the protocol headers contain:

 

The IEEE 802.11 MAC layer has several types of frames. The Data Frame is the frame type that contains the payload data, there is an ACK frame and RTS (Request to Send) an CTS (Clear to Send) frames. In addition there are Management Frame, Power-Save Poll Frame and Contention-Free Frame. With 802.11, all Data Frames gets an ACK receipt.

 

There are two types of PLCP headers. One for long preamble and one for short preamble. One downside to the PLCP is that the PLCP preamble and header is ALWAYS transmitted at 1Mbps, regardless of the transmit rate for the rest of the data. This means that the transfer time is constant at 192 Usec (microseconds) for the PLCP with long preamble. The short preamble version does a bit better, transmitting the shorter preamble at 1Mbps and the header at 2Mbps, shortening the transmit time to 96 Usec.

 

Transmitting the data is not just burping out a bunch of signals on the air. A strict set of rules is governing the way a transmission should behave (CSMA/CA). First the sender has to wait a period of DIFS (Distributed Coordination Function InterFrame Space) time of 50 Usec before the channel is presumed clear of traffic. Then a Data Frame or a Request to Send frame can be sent. The receivers answer to this is with an ACK or Clear To Send, accordingly. The receiver has to wait a SIFS (Short InterFrame Space) time of 10 Usec before this reply is sent.

This diagram shows the sequence of sending a typical TCP/IP datagram:

 

If the fragmentation threshold is set lower than the PSDU size or the TCP/IP datagram exceeds the maximum size of the PSDU, the data must be fragmented:

 

If RTS/CTS is employed, a RTS and CTS must be transmitted before any data can be transmitted:

 

These are typical settings from a Cisco AP 350 access point that can configure the fragmentation and RTS thresholds:

 

In the following calculation we have assumed that there are no collisions, no fragmentation, no RTS/CTS, no bursts and using long preamble. We also assume that each TCP/IP packet is given a TCP/IP ACK.

Mbit/s Net Mbit/s Efficiency
1 0.75 74.9%
2 1.41 70.7%
5.5 3.38 61.5%
11 5.32 48.4%

This shows that at the efficiency decreases as the wireless data rate increases and that this is much because of the PLCP.

When employing short preamble, this causes a significant effect:

Mbit/s Net Mbit/s Efficiency
1 0.77 76.9%
2 1.49 74.3%
5.5 3.83 69.6%
11 6.52 59.3%

Less time sending PLCP means more time for sending data.

 


Jardar.Leira@uninett.no 2005-04-15