Wireless Measurements lab 

Note: the alternative lab on router data analysis will be posted soon.

For this lab, you need to instrument the wireless network within a building of NYU. In Linux, you need to use "iwconfig" and "iwlist". In Windows, you can use a range of tools: "Netstumbler" is a good tool to measure signal strength characteristics and channel assignment.

1. Consider one of the three buildings (Warren Weaver Hall, Stern or Library). Using your wireless card, walk through every floor of the building to find out the list of access points that belong to a specific network (NYU-ROAM or Stern on the Move) and the channel assigned to each access point. Divide a floor into 9 regions (3x3 square) and approximate the location of each AP to the region where you observe the maximum signal strength. Note NYU-ROAM appears as a hidden channel with no SSID in Netstumbler or iwlist.

2. Pick 3 adjacent floors of the building and the access points installed in these floors. Determine the signal strength contour of each access point. How far can you see a single AP within a given floor and across floors? Describe the coverage, signal strength characteristics of each AP. Describe your inferences of these characteristics.

3. Signal strength vs performance measurement: Run an IPerf server on a machine that has a wired Ethernet connection to the Internet. Run an Iperf client on your laptop. From various points within the building (where one can get coverage to the same AP for different signal strengths), connect to the same AP and measure UDP available bandwidth, UDP loss rate and TCP bandwidth using IPerf. Compute the performance characteristics as a function of the signal strength of an AP and describe your inferences. 

Note: To install "iperf" at user-level, just download the tar.gz file into a linux machine and run "configure" and "make" (dont run "make install" - you need root access for that). After running "make", the executable will be present in src sub-directory.  Also there's a version of iperf for windows. Have the server version in a linux machine if possible.

Note: To connect to a specific AP in Windows, your card needs to recognize individual APs corresponding to each access network. If not, first use Netstumbler to detect APs corresponding to an SSID. If there exists only one, then connect to it assuming you have good signal strength. If you have two APs in vicinity for same SSID that your wireless connection device does not distinguish, then its hard to determine as to which AP it connects to. If you have only one, connect to that AP and keep walking away from it until you notice the signal strength to reduce and reach close to 30.  Till that point, typically your card is still associated with the first AP though other APs (for same SSID) might now be in vicinity. Once signal strength reaches below a certain threshold, the card shifts to a new AP if one is available. Using this crude approximation, you might still be able to do 3 & 4 though your card may not recognize individual APs belonging to same SSID. Otherwise, try to find other tools for windows or use Linux.

4. Simultaneous user experiment: Two users (two students) in parallel should connect to the same AP where one student has very high signal strength and the other student keeps roaming from  regions of high to low signal strength. Now simultaneously run TCP flows and UDP flows from both machines and measure the TCP and UDP throughput for the two machines. Compare the TCP and UDP throughputs of the two flows for varying signal strengths. Describe your inferences.

5. Channel conflicting regions: 802.11 uses only 3 non-overlapping channels (1,6,11) and many APs tend to have only one of these three channels. In general Channel X and Y overlap if their channel difference is less than 5. A location is said to be "channel conflicting" if 2 APs in the vicinity which have overlapping channels. Determine all conflicting regions within the building. Repeat this for multiple buildings (atleast for 2 different buildings). Is the channel allocation optimal? If not can you determine an alternate channel allocation which can minimize the number of conflicting regions?