Programming and Run-Time Systems for Distributed, Embedded and Handheld Systems



Principal Investigators: Zhiyuan Li, Saurabh Bagchi (ECE), Yung-Hsiang Lu (ECE), Lila Albin (Radiological and Environmental Management)

Current Participating Students : Matthew Tan Creti, Matt Beaman, Man Wang, Penxuan Zheng
Recent Ph.D's Graduated Cheng Wang (2004, now at Intel), Yonghua Ding (2004, now at IBM), Rong Xu (2005, now at HP), Changjiu Xian (2008, now at Microsoft)
Recent MS's Graduated Douglas Herbert

Current Sponsor: NSF/CNS/CT, NSF/CRI

This research project investigates operating system and compiler techniques for efficient and reliable use of memory, wireless network, and battery on wireless miniaturized computers, such as cell-phones, PDAs and sensor networks. Miniaturized computers are rapidly becoming highly popular devices for communication, computation, and environment monitoring and control. However, further development of applications for such devices faces a serious technical barrier due to the several resource constraints such as memory, wireless network bandwidth and battery capacity. The goal of the project is to find software solutions, based on wireless and distributed system technologies, to overcome such constraints. The current focus is on programming techniques for efficient error detection, diagnosis, and repair of system errors on sensor networks [IEEE SUTC06], [ ACM TAAS] .

Research has also addressed issues rising from the detection, diagnosis, and repair effort, including differential treatment of messages of different urgencies ( a paper for 27th IEEE International Symposium on Reliable Distributed Systems (SRDS) , Napoli, Italy, October 6-8, 2008.

The project team has also designed a multi-grade monitoring scheme to detect intrusions on pervasive computing systems. This new scheme addresses one of the most difficult problems for securing pervasive systems, namely, the powerful security measures invented previously for desktop machines and servers are too expensive to implement on the resource-constrained sensing and controlling devices often found in pervasive systems. There is not enough memory, processing capability, or battery energy to let the devices constantly run the intensive processing and communication operations. The multi-grade scheme uses multiple levels of security monitoring. It first uses resource-light monitoring to detect necessary conditions for malicious activities to occur. When a necessary condition is met, it then performs more expensive monitoring to detect sufficient conditions for malicious activities to occur [IEEE MASS 2009] .

On January 26, 2011, the project team has obtained an approval from NSF for a one-year no-cost extension. The main reason for this NCE request is due to recent changes happening to Crossbow, the manufacturer of the TelosB motes which are the most popular motes used in WSN research community. Crossbow has changed its product line and stopped producing TelosB or other motes. Several new vendors have emerged to compete for the market. Currently, it is not clear which vendors are reliable suppliers of TelosB or TelosB-compatible motes. The one-year extension will allow the team the time to investigate and identify a reliable supplier for the motes to complete the final phase of the development of the CO2 WSN proposed for this planning project. To enable diagnosis of the source of the detected errors, a dependence-based multi-level tracing and replay scheme is developed, which specifically targets resource-constrained distributed embedded systems such as wireless sensor networks. In the interest of portability across different hardware platforms, the scheme is implemented as a source-level tracing and replaying tool on top of the Open64 compiler which is tested using several TinyOS applications running on TelosB motes [ACM LCTES 2011] (Abstract and introduction).

A copy of the research prototype for the dependence-based record-replay tool is made available here for trials and comments by external users. Further improvements will be made over time, resource permitting.

In the past, we have designed compiler techniques to reduce memory consumption for programs which use arrays heavily. Some of these techniques are being applied to miniaturized computers. Recently, we have focused on studies of impact of various distributed software techniques on energy consumption on wireless-networked hendheld devices. Such studies have so far covered efficient main cache and mini cache use on Xscale-like processors [ACM/LCTES 2005] [IEEE/ISPASS 2004] [Journal on Embedded Computing, 2005] computation reuse [CGO04] , universal data-compression schemes [ICDCS03_extended] , IPSEC security protocol [WWC-5] [MM journal, 2004] , and computation offloading [ICPADS 2007], [PLDI 2004] . Recent research also covers issues of programming environment and task scheduling for low-power computing ([ISLPED07], [DAC07]). For computation offloading, we are treating it as a rather general approach to both memory-saving, speed improvement and battery-saving for miniaturized computers in a wireless LAN environment. We are building an infrastructure which integrates compiler analysis, run-time analysis and off-line profiling analysis to support the development of distributed applications and effective proxy service for wireless-networked devices [LCTES 2005] [PLDI 2004] (Abstract .pdf) [ICCCN 2004], [JPDC 2004] (.pdf) [SC03] (.pdf) [CASES01] (.ps) , [IPDPS02] (extended version)(.ps) .

Back to Li's home page