Wireless Data Throughput: Massimo Maffioletti Antonio Di Bacco Linkra Networks 12/04/2011
Wireless Data Throughput: Massimo Maffioletti Antonio Di Bacco Linkra Networks 12/04/2011
Antonio Di Bacco
White paper
LINKRA S.r.l. a socio unico Gruppo COMPEL ELECTRONICS S.p.A. C.F. / R.I. 02441450968 - VAT 02441450968 - R.E.A. MI 1455940 - Capitale Sociale 95.000,00 i. v.
ISO 9001-2000
White paper
The throughput specifications quoted can vary significantly depending upon which Layer is being considered.
FCS
IFG
PRE
MAC source
MAC source
Payload
FCS
IFG
PRE
Frame space (84 Bytes) Frame (64 Bytes) Packet (50 Bytes)
In the case of small frame sizes such as 64 bytes, the 20 bytes of preamble and IFG are significant compared to the frame. For a frame size of 1518 bytes, the preamble + IFG become less significant.
LINKRA S.r.l. a socio unico Gruppo COMPEL ELECTRONICS S.p.A. C.F. / R.I. 02441450968 - VAT 02441450968 - R.E.A. MI 1455940 - Capitale Sociale 95.000,00 i. v.
ISO 9001-2000
White paper
Layer 2 Throughput
Layer 2 throughput includes only the Ethernet frame, consisting of the payload, FCS and MAC Addresses, and excludes the preamble and IFG. In any given Ethernet system, the absolute maximum throughput will be equal to the data rate, e.g. 100 Mbit/s. In practice, these figures cannot be achieved because of the effect of frame size. The smaller size frames have a lower effective throughput than the larger sizes because of the addition of the pre-amble and the IFG bytes, which do not count as data throughput. The maximum achievable throughput for a 100 Mbit/s system for various frame sizes is given in Table 1. Table 1. Maximum Layer 2 Throughput for 100 Mbit/s Frame Size 64 128 256 512 1024 1280 1518 1522 Data Throughput Mbit/s 76.19 86.49 92.75 96.24 98.08 98.46 98.69 98.70 Preamble and IFG Mbit/s 23.81 13.51 7.25 3.76 1.92 1.54 1.30 1.30 Frames per Sec 148.809 84.459 45.289 23.496 11.973 9.615 8.127 8.106
In radio systems with integrated Layer 2 switching, like Linkras radio systems, pre-amble and IFG data is stripped from the incoming data stream and, instead of this header, a shorter trail sequence is added and transmitted over the radio link. At the far end network interface, the preamble and IFG bytes are re-inserted into the data stream by the radio equipment.
Layer 1 Throughput
By quoting throughputs at Layer 1, vendors intentionally include preamble and IFG in their figures, leading to inflated results for smaller frame sizes. It represents a classic example of how specsmanship is employed to pump up performance figures. Figure 2 below shows the difference between claimed Layer 1 throughput from a leading radio vendor, and the measured Layer 2 throughput. For larger frame sizes the difference is small, but for 64 byte frames the difference is enormous a measured Layer 2 throughput of 30.4 Mbit/s is actually advertised as being 40 Mbit/s! Figure 2. Claimed Layer 1 versus Measured Layer 2 Throughput
Throughput
1024
1280
1512
LINKRA S.r.l. a socio unico Gruppo COMPEL ELECTRONICS S.p.A. C.F. / R.I. 02441450968 - VAT 02441450968 - R.E.A. MI 1455940 - Capitale Sociale 95.000,00 i. v.
ISO 9001-2000
White paper
To further compound the spin, vendors then quote this Layer 1, 64 bytes frame size throughput (rounded up of course) as being the maximum platform capacity, ie: 40 Mbit/s, with no qualification. You may even see figures as high as 40 Mbit/s, which is achieved by adding in the radio overhead for FEC, management and aux channels, giving the additional 10 Mbit/s, all with the aim to make the quoted figure as high as possible. Who knows, maybe soon we will see claims of 50 Mbit/s! RFC 2544 was used for the measured Layer 2 throughput in Figure 2 above. Linkra uses RFC 2544 testing to validate the performance of all its Ethernet products.
Net MAC rate available(Mbps) versus layer 1 throughput @frame size=64 and 1500 bytes
450 400 350 300 250 200 150 100 50 0 4QAM 16QAM 64QAM 256QAM
L2 56MHz@64 L1 56MHz@64 L2 56MHz@1500 L1 56Mhz @1500
LINKRA S.r.l. a socio unico Gruppo COMPEL ELECTRONICS S.p.A. C.F. / R.I. 02441450968 - VAT 02441450968 - R.E.A. MI 1455940 - Capitale Sociale 95.000,00 i. v.
ISO 9001-2000
White paper
Higher Throughputs
So now, some vendors are claiming 800, and even 900 Mbit/s throughputs. How do they achieve that? The secret here is that this requires two radio links running in parallel, effectively doubling the link capacity. This is a standard microwave technique and is made even more effective utilizing radio techniques such as XPIC (Cross Pol Interference Cancellation). However, the network interface provided usually consists of two GigE interfaces (one each 400/450 Mbit/s RF link). These radios usually require an external Ethernet switch to combine the traffic from both links into a single virtual high capacity link.
Summary
Some vendors inflate the actual throughput over a radio link by as much as 20-30%, even though all radios employ similar methods to package and transport data. Physics dictates that there is no free lunch, meaning that if someone is claiming dramatically better performance then you should really check the fine print!
5.
Bibliography
Benchmarking Methodology for Network Interconnect Devices RFC 2544 RFC 2544 Testing of Ethernet Services in Telecom Networks Agilent Technologies Benchmarking wireless ethernet throughput Arris-Stratex 1-30-2008
LINKRA S.r.l. a socio unico Gruppo COMPEL ELECTRONICS S.p.A. C.F. / R.I. 02441450968 - VAT 02441450968 - R.E.A. MI 1455940 - Capitale Sociale 95.000,00 i. v.
ISO 9001-2000