Aleksandar Milenkovic - US grants

The next generation of wireless is in distributed sensor networks. Applications in this area are a rapidly growing research area. The project will investigate possibilities for coordinated action to improve the energy efficiency in distributed sensor network on all levels, from software to component design to system integration. The overall goal and impact of this will be the decrease in the cost of sensor network, allowing wider, more capable deployments for generalized information gathering and surveillance. A test-bed environment for energy-efficiency and performance evaluation of reconfigurable hierarchical networks of sensors will be established at the Center for Nanosensor Technology at the University of Alaska Fairbanks. Another anticipated benefit of such intelligent surveillance sensor networks is the extension of the same principles to everyday applications of distributed sensor networks.

A recent shift from single-core to multi-core and many-core architectures and rising complexity of both hardware and software pose a number of challenges to computer systems industry. Computer architecture research is crucial in finding new approaches for designing, programming, debugging, and operating future computer systems based on multi-core processors.
Computer architecture research has predominantly relied on software simulations for quantitative evaluations of new architectural ideas and design space exploration in the last two decades. However, simulation-based research cannot keep pace with growing demands of new architectures and new benchmarks, suffering from extremely long simulation times, over-burdening complexity, and limited credibility. FPGA-based hardware emulators provide an attractive and cost-effective alternative to simulation: they address key weaknesses of software simulators - scalability, speed, and credibility, while offering similar levels of flexibility, observability, and reproducibility. Modern FPGAs can accommodate up to 16 processor cores running at clock speeds of over 200 MHz. At these speeds, they enable full-system emulation of complex multi-core architectures, run operating systems and real-world applications, and are several orders of magnitude faster than corresponding simulators.
This project will acquire new FPGA-based emulation hardware infrastructure at the University of Alabama in Huntsville. It will be used by researchers in the Laboratory for Advanced Computer Architectures and Systems (LaCASA), to help their ongoing research efforts targeting new architectures for increasing programmers? productivity and dependability in both single-core and multi-core computer systems. The infrastructure will also be used in several graduate and undergraduate computer engineering courses to improve academic training and learning experience, and help recruiting efforts.

Embedded computer systems have become essential to many aspects of our lives. Cheaper, smaller, faster, more sophisticated, and more energy-efficient embedded devices spur ever new applications. However, the growing complexity and shift to multicores make programming and debugging these systems difficult. Traditional debugging is time consuming and may interfere with program execution, causing some bugs to become irreproducible and making it unusable in real-time environments. Moreover, tracing a processor?s internal state during execution is only feasible for short program segments and requires large on-chip buffers or wide trace ports, either of which increases system cost and limits scalability. This project is developing the next generation of trace compression methods and infrastructure to make continuous, real-time, unobtrusive, and cost-effective program, data, and bus tracing possible in embedded systems. The approach relies on on-chip hardware to record the processor state and corresponding software modules in the debugger. The novel insight is that a sequence of trace records can be translated, without loss of information, into a much shorter sequence of miss events using small hardware structures. The few remaining miss events are then further compressed using highly-effective yet simple-to-implement encoding schemes, yielding heretofore unseen compression ratios.

The new tracing and debugging infrastructure can help programmers find difficult and intermittent software bugs faster, thus improving productivity. For example, reducing debugging time by just one percent amounts to hundreds of millions of dollars annually in saved salaries, with a concomitant reduction in software cost and time to market. Moreover, higher quality software may eliminate errors in medical, automotive, or mission-critical devices and thus save lives.

Healthcare systems are facing an imminent crisis precipitated by the confluence of current social, economic and demographic trends. According to the United States Census Bureau our total expenditures on healthcare reached $2.5 trillion in 2009 and are projected to reach 20% of the nation's Gross Domestic Product in 2020. These statistics suggest the need to seek more scalable and affordable healthcare solutions. A focus on proactive management of wellness / prevention, on early detection of disease, and on optimal maintenance of chronic conditions has the potential to reduce healthcare costs while increasing quality of life. Recent technological advances in sensors, integration and miniaturization of low-power electronics, wireless networking, mobile computing, and cloud computing allow us to modernize and change the way health care services are deployed and delivered.

Mobile health monitoring systems that integrate wireless body area networks (WBANs) with a range of intelligent and miniature sensors, personal devices like smart phones, and servers that can be accessed over the Internet emerge as a promising technology for real-time, unobtrusive health and wellness monitoring of individuals during normal daily activities. Unprecedented proliferation of personal computing devices and smart sensors makes such solutions practical and affordable. But further research is needed to explore the design space and create optimal solution for a given application, and also to understand the full scope of the opportunities offered by the new technology while addressing the challenges.

This project will provide funds to create mHealth, a new computing infrastructure to support research and education in computer systems for mobile health and wellness monitoring at the University of Alabama in Huntsville. The mHealth infrastructure consists of: a) A variety of wearable wireless sensors for monitoring users' physiological signals, body movement and activity levels, along with general environmental conditions; b) Personal devices that collect data from the sensors, analyze it, compile personalized health status information, and upload data over the Internet to a server; and c) An mHealth server running databases and services for logging and analysis of health records from multiple users. The mHealth infrastructure will directly support several research efforts conducted by the investigators, their students, and their collaborators at the University of Alabama in Huntsville's College of Nursing, at the Mayo Clinic, at the University of Alabama at Birmingham's School of Health Sciences, and at the Lakeshore Foundation in Birmingham. In addition, mHealth will be used in a senior design course sequence and two graduate courses at the University of Alabama in Huntsville to improve the student academic experience and to help with recruiting efforts through outreach activities.

Intellectual Merit
The mHealth infrastructure will enable the investigators to pursue the following research goals: 1) Exploring critical design issues in the next generation of wireless wearable body area networks for health monitoring including their functionality, reliability, and energy-efficiency; 2) Creating annotated public data repositories with vital signs and physical activity parameters during normal daily activities to promote further research in WBANs for health monitoring; 3) Building WBAN research prototypes and developing algorithms, firmware and software artifacts for applications such as monitoring and managing physical activity in people with spinal cord injury (with UAB, Lakeshore), monitoring and managing ambulatory rehabilitation of people with coronary disease (with Mayo), health status assessment of people with coronary disease (with Mayo), and monitoring of the occupational stress of nurses (with UAHuntsville Nursing).

Broader Impact
Acquisition of the mHealth infrastructure will help the investigators and their collaborators pursue research in several multidisciplinary areas. If successful, these efforts promise to improve our understanding of the design space of wearable ubiquitous platforms for health monitoring, to improve quality of life of people with disabilities, to promote healthy behaviors through use of technology in the general population, to enable affordable solutions for ambulatory rehabilitation, and to enable early detection of health conditions. The mHealth infrastructure will also be used directly for education and training in several graduate and undergraduate courses.

Non-volatile flash memories, the basic building blocks of solid-state storage devices, offer small form factors, high-capacity, high-speed, and low-power permanent storage solutions in a wide range of computing systems found in consumer electronics, automotive, military, industrial, healthcare, and enterprise segments. Unfortunately, deleting data instantly from physical flash memories is not always straightforward as it incurs hefty overheads and increases wear level, especially in solid-state drives (SSDs). According to a recent report, 42% of used SSDs sold on eBay held sensitive recoverable data, even though data deletion or sanitization methods were in place. Whereas standard data deletion methods make the data inaccessible to the user through standard interfaces, our recent research efforts demonstrate that the deleted data is partially or fully recoverable by means of physical characterization of flash memory cells. This underscores the need for finding new ways to ensure that deleted data is promptly, permanently, and irreversibly re-moved from flash memories.

This project is investigating new cost-effective instant data sanitization techniques for flash memories. The proposed techniques take advantage of (a) timeline of recent flash operations; (b) physical properties of flash memory cells cap-tured by state decay models and (c) partial flash program and erase operations. These techniques will be tailored to uti-lize the unique properties and architectures of flash-memory chips found in commercial storage applications. The proposed techniques will be applicable to different types of flash memories, will require no or minimal changes in hardware, and will not increase the wear level of flash memories nor significantly increase the latency of common flash operations. Techniques for cost-effective data sanitization of flash memories in run-time will benefit consumers, industry, and government alike by ensuring that deleted data is not recoverable at any time during the product?s life cycle. These techniques will be deployed in the firmware of flash memory controllers. A direct outcome of this proposal will be the training of three graduate students in the important area of hardware-oriented security and memory systems. The investigators are developing teaching materials to introduce hardware-oriented security topics in embedded systems, computer systems architecture, and hardware reliability courses.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.