CS 536 (Park) Project Ideas Nov. 8 (Mon.), 2004Submission instructions: The project output, including software, electronic report, and hardcopy report, are due onDec. 18, 2004 (5pm). The electronic material should be submitted using turnin. The hardcopy can be slid under myoffice door (CS220). Other project related specifications are given in Problem 3 of Assignment VI. The project topicsneed not be confined to the ones listed below. The programs may be implemented in C, C++, perl, and shell scripts(or a combination). Java and other high-level languages are not allowed.Topic 1: Wireless Ad Hoc ForwardingSuppose we have a number of wireless stations communicating with each other in ad hoc mode, where the aim isto send files amongst each other. For concreteness, assume there are three stations A, B, and C (one laptop andtwo pocket PCs). A and C may be too far apart for IEEE 802.11 to successfully transmit data, but if B is wedgedin-between then A may use B as a relay station to reach C (and vice versa). The aim is to build a wireless filetransfer client/server application (call it wftp) that uses sensing and tunneling to achieve this goal. One may assumethat initially a file containing the IP addresses of all ad hoc wireless stations are given in a file at all stations (howit gets there is a bootstrap problem). A, to send a file to C, would execute% wftp IP-address file-namewith C’s IP address. The file transfer application at A would first check if C’s address is valid (a member of the adhoc group) and directly reachable from A. If so, it would directly transmit to the file transfer server process, wftpd,running on C. If not, it has to first relay through B, which means tunneling through B. That is, A’s wftp sendsto B’s wftpd which, in turn, forwards to C’s w ftpd. Since A may have other potential relays besides B in thesame ad hoc group, it may first send a query message to the potential relays one-by-one, receiving both reachabilityand performance related information from the candidate relays. The performance related information may be delayvalues from ping or signal strength values obtained from the WLAN driver (iwconfig is the command-line wirelessextension to ifconfig). Benchmarking entails using 3–4 wireless stations that are placed in a line or diamond-likeconfiguration so that relaying is necessary, then measuring the throughput achieved by the wftp application throughclever choice of forwarding. Baseline benchmarks include throughput tests, as a function of distance, between twodirectly reachable stations. Then the overhead and slow-down stemming from forwarding when a relay station iswedged in-between. Since TCP ACK traffic—in the opposite direction—may collide with TCP data traffic at therelay station (an instance of the “hidden station” problem), other interesting benchmark variations are to comparethroughput, under the same set-up, with/without RTS/CTS turned on which is a simple matter of choosing theRTS frame threshold. wftp should work correctly: files are transported with no corruption or missing pieces. Theinteresting question is: how well does it do its job. Laptops and pocket PCs may be borrowed for benchmarkingfrom my lab. Most of the code development can be done in Xinu, except if signal strength is to be used directlyfor next hop selection. Porting C/C++ code to pocket PCs (they run Familiar Linux) requires cross-compilation atopen sites operated by handhelds.org (supported by Compaq/HP), a straightforward matter.Topic 2: WLAN Handoff PerformanceMobility support through WLAN handoff is achieved by mobile host/AP reassociation in conjunction with wirelineEthernet frame forwarding via bridging. A potentially interesting question is: what impact does handoff over-head between APs exert on UDP-based real-time streaming performance and TCP-based file transfer performance?For research purposes, we operate a 6-AP 802.11b WLAN testbed (SSID “Park”) in the CS Building (2 APs perfloor) that are connected through an Ethernet switch. Performance evaluation would entail running a UDP CBRsender/receiver in close proximity under a given AP, then move the sender (or receiver) around the 3 floors fromAP to AP triggering reassociations. From the logged throughput measurements—including jitter in the delay timestamps—and hand-logged (approximate) position in the CS Building the magnitude of the impact of handoff on CBRtraffic (representative of VoIP) may be discerned. The same goes for TCP-based file transfer where backoffs may betriggered during handoff which have performance repercussions after the functional handoff is completed. Subtletiessuch as, “does a noticeable/reliable performance degradation materialize prior to handoff as a mobile approaches a‘grey’ zone?” pose interesting questions to understanding how bad or good the current handoff scheme is, and whereit may be lacking. Associations can be tracked at the APs using SNMP logs/probes that give a precise indication ofhandoff and resultant performance.Topic 3: The Shape of Hot Spot TrafficUsing a powerful WLAN sniffer that records all relevant performance metrics—number of active wireless stations,their LAN throughput, per packet/frame signal strength, error rate, payload type, payload size, nested payloads re-quiring deep packet inspection, etc.—one may evaluate the static and dynamic make-up of WLAN Internet traffic ata popular hot spoterror rate, payload type, payload size, nested payloads requiring deep packet inspection, etc.—onemay evaluate the static and dynamic make-up of WLAN Internet traffic at a popular hot spot such as the VillageCoffee Shop. Of course, any non-protocol related user payload information must be discarded. From the log files,the aim is to do a careful analysis of the data that exceeds the sometimes coarse conclusions that such tools alreadycarry out. In the static case, statistics on what fraction of traffic is UDP, TCP, and application layer protocols aboveUDP/TCP (such as HTTP), what is the distribution of frame/packet sizes, what is the lifetime distribution of TCPsessions, how many bytes do their transfer (“elephants and mice”), what is the throughput of the WLAN, what isthe throughput share of individual wireless stations, and so forth are stats relevant for future engineering of WLANInternet access networks. On the dynamic front, what’s the pattern of TCP session arrivals over time, what’s thepattern of UDP flow arrivals over time,