Conclusion


нннннн Wireless LAN - conclusion

This section will sum up the results of our research into wireless LAN. The research was done by a collaboration of ACCU, Digital, Lucent, Surfnet and the computational physics workgroup of Utrecht University. One of the aims of the project is to advise the board of the University College on the suitability of a wireless LAN for the College campus. In the near future, the focus of the research will shift to simulating the MAC layer behaviour of the 802.11 wireless LAN standard. This will allow us to gain a more fundamental insight in the possibilities and restrictions of wireless LAN.

Scaling

One of the major concerns about wireless LAN was whether or not the throughput would scale with the number of clients. Under extreme loads, collissions are very likely to occur. At a certain collission rate, the system will drown in retries, that in their turn collide and generate a retry. This effect, that is known to occur in ethernet, would render a very low throughput in situations where you need the most.
In order to investigate this effect, a throughput measurement tool was developed in our workgroup and two test sessions were organised in November and December. It allows to have clients sending as much garbage data as they can to a server application, which calculates the throughput. The aggregate throughput of up to 25 clients was seen to remain stable at 1.5-1.8 Mbit/second, depending on block size. These results are backed by independent measurement using existing tools. Further, two operating system/protocol combinations (TCP-IP under Windows and IPX under Novell) were seen to produce similar results. Therefore, we conclude that wireless LAN is able to deliver reliable communication for a large number of users. Detailed testplans and results of the two scaling tests are published here;
Testplan and results of the November test.
Testplan and results of the December test.

Multicast

For future applications in education multicast might prove to be very useful to reduce the necessary bandwidth. Therefore, it is important that the wireless LAN truly supports multicast. We tested this feature by having one PC broadcast a video to three clients via an access point. The clients each received up to 1.3 Mbit/s. This is only possible using multicast, as the aggregate throughput would otherwise be nearly 4 Mbit/s.
 

Architecture

According to the 802.11 standard, the frequency band from 2.400 to 2.483 GHz has been assigned for wireless LAN use world wide. It is possible to operate more than one overlapping cell at different frequencies. However, if the frequencies are too close, interference will distort both signals and lead to much lower throughputs.н Thus, the placement of the access point on the campus determines the network capacity in each area.
In order to estimate the number of access points needed and the possibility to raise the bandwidth in designated areas, a number of tests were performed. A first test consisted of throughput measurements for clients communicating to two access points at different frequencies in the same room. It showed that the frequency difference should be at least 30 and preferably 40 MHz to avoid interference. With the product under investigation, two overlapping cells are allowed. When the full band will be used it should be possible to operate three independent cells, yielding a total throughput of 24 Mbit/s (Later this year, new cards with up to 8 Mbit/s will be launched).
On the other hand, the range of each access point is an important parameter. It depends mainly on the amount and sort of material obstructing the signal and might differ significantly from building to building. Therefore, the quality of the signal was measured in a future University College building, having placed one access point. Horizontally, the range was found to be of the order of 50 meters. Vertically, good signal is basicly confined to one floor. In most areas, the range of the signal is not going to be the limiting factor, as too many users in one cell will result in low service speed.
The results mentioned in this section are published here;
The range measurements.
The interference measurement.

Roaming

For three months, a wireless LAN has been used as production network at two locations in Utrecht University. During this period, it has proved to be a stable, easy-to-manage and mobile networking solution. Roaming within one subnet is secure and fast - the connection being lost for milliseconds. For a change of frequency, a simple user intervention is still required. If one wants to allow roaming between subnets, mobile IP will have to be implemented.

Conclusion

Generally, wireless LAN proved to be a reliable and reasonably fast mobile networking solution. For most purposes, the bandwidth it provides should suffice, although cabled LANs offer more possibilities to extend the bandwidth. In situations where mobility is as good a criterium as bandwidth, a wireless LAN should certainly be considered a serious option.

dr.ir. C.Th.A.M. de Laat, Faculty of Physics and Astronomy, Utrecht University, Princetonplein 5, NL-3584CC Utrecht, The Netherlands.
Tel: (31)30-2534585, Fax:(31)30-2537555
CdL -  27 dec 1997  Visitors since 27-12-1997: [an error occurred while processing this directive] M.Vos
Backdrop pictures are Copyright Morris&Goscinny, dargaud, dupuis.