UCSD CSE accomplishments of research on resource management in HPWREN: QoS scheduling, routing and power management

The research focus of a UCSD Computer Science and Engineering team, which is collaborating with the HPWREN project, has been on providing QoS to various applications in HPWREN while minimizing the power consumption of battery operated wireless nodes. During this last period we focused on QoS network scheduling and power management. The result of our research are three journal papers, four conference papers, 11 invited talks in industry, conferences and abroad, approximately $400k of funding obtained from industry and technology transfer to Sun Microsystems, Qualcomm and Cisco. One MS thesis has already been completed, while another MS and PhD are in progress. Two students have been funded directly through HPWREN funds, and two other students have been involved through industry funding we have been able to obtain.

The about 4500 acre San Diego State University's Santa Margarita Ecological Reserve has been chosen as the initial project testbed for the UC San Diego research. The geographic layout is well suited for QoS research on access networks, which can aggregate traffic from many sensors and forward to the HPWREN backbone.

Our distributed scheduling algorithm reduces the interference of active wireless nodes and limits contention, while decreasing the power consumption. The simulation and measurements performed with data we gathered on HPWREN access networks show that with our scheduling algorithm it is possible to achieve a throughput improvement of up to 10% and maximum power saving of 85%, while providing more predictable delay. We can provide even larger power savings on each node by implementing our novel online learning algorithm to drive system level power management. We formulate both dynamic power management (DPM) and dynamic voltage frequency scaling (DVFS) problems as one of workload characterization and selection, and solve it using this algorithm. Our results show that the algorithm adapts really well with changing device and workload characteristics and achieves an overall performance comparable to the best performing expert at any point in time. Moreover, it is extremely lightweight and has almost negligible overhead. For more information on our work, please see the published papers listed at the end of this short report.

In the next period we will focus on energy efficient and QoS aware routing methodologies. Preliminary studies of QoS routing algorithms have already been done during this last year, along with an initial implementation of routing functionality in Cisco routers used at HPWREN backbone. During this year we plan to expand this work to include more sophisticated routing solutions, and to expand so that routing can be implemented in energy efficient manner on wireless nodes.

Journal Papers:

1. E. Regini, D. Lim, T. Simunic Rosing, "Distributed scheduling for heterogeneous wireless sensor networks," submitted to IEEE Transactions on Mobile Computing, 2007.
2. G. Dhiman, T. Simunic Rosing, "Using online learning for system level power management," submitted to IEEE Transactions on CAD, 2007.
3. J. Kim, T. Simunic Rosing, "Power-aware resource management techniques for low-power embedded systems," in Handbook of Real-Time and Embedded Systems, Edited by S. H. Son, I. Lee, J. Y-T Leung, Taylor-Francis Group LLC, 2006.

Conference Papers:

1. E. Regini, D. Lim, T. Simunic Rosing, "Heterogeneous wireless network management," Submitted to INFOCOM'08.
2. G. Dhiman, T. Simunic Rosing, "Dynamic voltage scaling using machine learning," ISLPED'07.
3. D. Lim, J. Shim, T. Simunic Rosing, T. Javidi, "Scheduling data delivery in heterogeneous wireless sensor networks," ISM'06.
4. G. Dhiman, T. Simunic Rosing, "Dynamic power management using machine learning," ICCAD'06

Thesis:

1. D. Lim, "Heterogeneous wireless network management," MS Thesis, UCSD, 2006.
2. E. Regini, "Distributed network management and scheduling, " MS Thesis, UCSD, In progress.
3. G. Dhiman, "Routing under energy and QoS constraints in heterogeneous wireless sensor networks, " PhD Thesis, UCSD, In Progress.

Invited talks:

1. Intel, Hillsboro, OR: "Heterogeneous wireless network management," July 2007.
2. Intel, Hillsboro, OR: "Using machine learning for power management in wireless networks," July 2007.
3. SECON, June 2007: "Heterogeneous wireless sensor network resource management," Invited talk.
4. Qualcomm, San Diego, CA: "HPWREN - challenges and research opportunities," June 2007.
5. Qualcomm, San Diego, CA: "Distributed scheduling in wireless sensor networks, " June, 2007.
6. Sun Microsystems, Menlo Park, CA: "Heterogeneous wireless network management," December 2006.
7. Sun Microsystems, Menlo Park, CA: "Using machine learning for power management in wireless networks," December 2006.
8. HP Labs, Palo Alto, CA, ""Resource Management of Heterogeneous Wireless Embedded Systems", 2006.
9. Ericsson, San Diego, CA: "Heterogeneous sensor networks," 2006.
10. IMEC, Belgium, "Resource Management of Heterogeneous Wireless Embedded Systems", 2005
11. DATE'05, Munich, Germany, Special Session Invited paper: "Power management in wireless environments," 2005.

Technology transfer:

1. Qualcomm, San Diego, CA, summer 2007: student (Edorado Regini) as an intern implements power management for Qualcomm products
2. Sun Microsystems, Menlo Park, CA, summer 2007: student (Gaurav Dhiman) as an intern works on network resource management implementation in Solaris kernel networking code
3. Cisco, San Jose, CA, summer 2007: student (Ben Lee) works on routing algorithm implementations.
4. Samsung, January 2006: student (Daeseob Lim) implements power management for Samsung's products.

Industry funding:

1. Sun Microsystems, 2007 - $58k for research in network routing and scheduling
2. HP, Qualcomm and Sun Microsystems through CNS, 2006 - $130k for research in wireless sensor network resource management
3. HP, Qualcomm and Sun Microsystems through CNS, 2005 - $60k for research in heterogeneous network management
4. Los Alamos National Lab, 2006, 2007 - $85k for research in power management in wireless sensor networks

Prof. Tajana Simunic Rosing
Department of Computer Science and Engineering
University of California San Diego