Jingpu Shi Department of Electrical and Computer Engineering, Rice University 6100 Main Street, MS-366, Houston, TX 77005 Phone: 713-348-3776, Email: jingpu@rice.edu OBJECTIVE To obtain a position in wireless networking. EDUCATION Rice University, Houston, TX PhD Graduate Student, Electrical and Computer Engineering 08/2004-Present The Ohio State University, Columbus, OH MS, Electrical Engineering 08/2004 University of Electronic Science and Technology, China BS, Electrical Engineering 07/1996 TEACHING EXPERIENCE Held office hours, advised students and occasionally gave lectures for course: ELEC 437, Introduction to Communication Networks COMP 429, Introduction to Computer Networks ELEC 537, Communication Networks ELEC 438, Wireless Networking for Under-Resourced Urban Communities Fall 2005 Spring 06 Fall 2006 Spring 2007 RESEARCH EXPERIENCE Research Assistant 08/2004-Present Rice University, Houston, TX, Department of Electrical and Computer Engineering Investigated starvation in operational urban mesh networks. Modeled synchronized contention in CSMA multi-hop networks. Designed a multi-channel MAC protocol to address starvation in multi-hop networks. Analyzed and modeled medium access in CSMA multi-hop wireless networks. PUBLICATIONS 1. J. Shi, T. Salonidis and E. Knightly, "Modeling Fairness And Clock Drifts Under Synchronized CSMA Contention," to be submitted. 2. J. Shi, O. Gurewitz, V. Mancuso, J. Camp and E. Knightly "Starvation in operational urban mesh networks: origins and solutions ", submitted to SIGCOMM 2007. 3. J. Shi, T. Salonidis and E. Knightly, "Starvation Mitigation Through Multi-Channel Coordination in CSMA-based Wireless Networks," in Proceedings of ACM MobiHoc'06, Florence, Italy, May 2006 (acceptance rate 10%). 1 4. M. Garetto, J. Shi and E. Knightly, "Modeling Media Access in Embedded Two-Flow Topologies of Multi-hop Wireless Networks," in Proceedings of ACM MobiCom'05, Cologne, Germany, August 2005 (acceptance rate 10%). . SKILLS NS-2 network simulator, C, C++, AWK, TCL, MATLAB, Unix, Linux, Windows. SERVICES, SCHOLARSHIPS AND AWARDS Rice University - Rice University Fellowship 2004-2005 The Ohio State University - Outstanding Graduate Student Award (one of the three recipients) - President of Cultural Ambassadors Association - Voting member of Academic Misconduct Committee - Engineering school representative in Research and Graduate Council - Department representative at Council of Graduate Students - University Fellowship RESEARCH PROJECTS Rice University, Houston, TX Department of Electrical and Computer Engineering 2003 2003 2003 2003 2002 2001-2002 08/2004-Present Countering starvation in operational urban networks Tree-structured mesh networks that employ IEEE 802.11 lead to acute unfairness or starvation, even for flows under end-to-end congestion control. Namely, mesh nodes closer to the wired gateway receive significantly higher throughput than mesh nodes farther away. I first present measurements obtained in a Houston urban network to show severe throughput imbalance among TCP flows. To address this problem, I identify a simple scenario as the critical building block for understanding and predicting the unfairness behavior observed by TCP flows traveling over a mesh network. I propose a discrete time Markov chain model to investigate the interaction between end-to-end congestion control and the MAC congestion window. The model reveals the key role of the last hop in controlling the stationary distribution of the system. Consequently I develop a simple contention window policy to allow fair gateway bandwidth sharing between the mesh nodes. I finally implement this policy in hardware and validate its effectiveness through on-site deployment. 2 Model and Fairness properties in synchronized CSMA wireless networks A broad class of CSMA protocols use synchronized contention in which nodes periodically contend at intervals of fixed duration. While synchronized CSMA contention has not been proposed to improve fairness, in this project I show that it has the potential to address the severe throughput imbalances that plague asynchronous CSMA protocols such as IEEE 802.11 DCF. However, its fairness properties have never been analyzed. I introduce an analytical model that directly quantifies the interplay and impact of the main performance factors: imperfect clock synchronization, contention window size, carrier sense, topology properties and usage of guard time. This model reveals conditions on protocol parameters under which the throughput of certain flows can exponentially decrease with clock drift. In addition, it enables solutions that can offset such problems in a predictable manner. Starvation Mitigation Through Multi-channel Coordination in CSMA-based Wireless Networks Existing multi-channel protocols have been demonstrated to significantly increase aggregate throughput compared to single-channel protocols. However, I show that despite such improvements in aggregate throughput, existing protocols can lead to flow starvation in a multi-hop network. In this project, I devise Asynchronous Multi-channel Coordination Protocol (AMCP), a distributed medium access protocol that not only increases aggregate throughput, but, more importantly, addresses the fundamental coordination problems that lead to starvation. I analytically derive and experimentally validate an approximate lower bound on the throughput of any flow in an arbitrary topology. I also demonstrate that AMCP can deliver significantly higher per-flow throughput than both IEEE 802.11 and existing multi-channel solutions. Modeling Media Access in Embedded Two-Flow Topologies of Multi-hop Wireless Networks I decompose a multi-hop wireless network into embedded subgraphs, each consisting of four nodes and two flow pairs. I identify all twelve possible topologies that arise according to whether the different nodes are in radio range of each other. I show that under both a random spatial distribution of nodes and random waypoint mobility with shortest-path routing, a critical and highly probable scenario is a class in which the channel state shared by the two flows is not only incomplete (i.e., the graph is not fully connected), but there is also asymmetry in the state between the two flows. I develop an accurate analytical model validated by simulations to characterize the long-term unfairness that naturally arises when CSMA with two or four-way handshake is employed as a random access protocol. Moreover, we show that another key class of topologies consists of incomplete but symmetric shared state. I show that in this case, the system achieves long-term fairness, yet endures significant durations in which one flow dominates channel access with many repeated transmissions before relinquishing the channel. 3 SIMULATIONS AND IMPLEMENTATIONS Rice University, Houston, TX Department of Electrical and Computer Engineering 08/2004-Present Synchronized CSMA (S-CSMA) S-CSMA is a single-channel synchronized CSMA protocol in which time is partitioned into fixed-duration cycles. In a perfectly synchronized system, each node begins each cycle at the same time instant. Each cycle consists of a contention phase followed by a data transmission phase. At the beginning of each cycle, each node senses the medium and waits while it is busy. The data transmission phase may consist of one or more data-link frames and their corresponding acknowledgments provided that all transmission (data and acknowledgment) is completed by the end of the cycle. I implemented S-CSMA using ns-2 (ns-2.28) simulator with CMU wireless extensions. I wrote scripts and evaluated its fairness properties in various experiments under both perfect and imperfect synchronization. Asynchronous Multi-channel Coordination Protocol (AMCP) AMCP is a MAC protocol that utilizes multiple channels to increase per-flow throughput. It uses a dedicated control channel on which nodes contend to reserve data channels by exchanging RTS/CTS packets according to 802.11 DCF. All channels are orthogonal with respect to each other. Each node has a single transceiver, hence it can either transmit or listen, but not both. Also it can listen to or transmit on one channel at a time. To execute AMCP, each node maintains a local channel status table. I implemented AMCP using ns-2 (ns-2.28) simulator with CMU wireless extensions. I developed scripts and evaluated AMCP in both single-hop and multi-hop topologies. Multi-band Opportunistic Auto Rate (MOAR) Multi-band Opportunistic Auto Rate (MOAR) is an opportunistic media access protocol for multi-rate IEEE 802.11. Building on the IEEE 802.11 MAC protocol, MOAR uses an optimal band skipping rule to find the best band for transmission every time a node pair gains access to the medium. I implemented this protocol to systematically investigated its performance in various network settings. I utilized Opportunistic Auto Rate (OAR) as a starting point. The NS extensions include Rice Networks Group (RNG) implementation of OAR for ns-2.1b7. The implementation of Ricean Fading Model is based on CMU additions to NS to handle Ricean and Rayleigh fading. All simulation results are presented in the following paper: A. Sabharwal, A. Khoshnevis and E. Knightly, Opportunistic Spectral Usage: Bounds and a Multi-band CSMA/CA Protocol, to appear in ACM Transactions on Networking, 2007. 4 REFERENCES Professor Edward W. Knightly Department of Electrical and Computer Engineering Rice University 6100 Main Street, MS 380 Houston, TX 77005-1892 (713) 348-5748 knightly@ece.rice.edu Professor Ashutosh Sabharwal Department of Electrical and Computer Engineering Rice University 6100 Main Street, MS 380 Houston, TX 77005-1892 (713) 348-5057 ashu@ece.rice.edu Professor David B. Johnson Department of Computer Science Rice University 6100 Main Street, MS 132 Houston, TX 77005-1892 (713) 348-3063 dbj@cs.rice.edu Dr. Omer Gurewitz Department of Electrical and Computer Engineering Rice University 6100 Main Street, MS 380 Houston, TX 77005-1892 (713) 348-2359 gurewitz@ece.rice.edu 5