UT Dallas Computer Science professor and expert in the field of Real-Time Systems, Dr. Cong Liu, recently had three papers accepted into the 2016 IEEE Real-Time Systems Symposium (RTSS). Two of the papers that were accepted had Zheng Dong, his PhD student, as the first author. Each year RTSS accepts between 20 to 30 papers, with an acceptance ratio around 20%.
The IEEE Real-Time Systems Symposium (RTSS) is the premier conference in the area of real time systems where members of both academia and industry come together to present cutting-edge research in the field of real-time systems with respect to both theory and practice. RTSS provides a forum for the presentation of high-quality, original research covering all aspects of real-time systems theory, design, analysis, implementation, evaluation, and experiences. The symposium is continuously looking to embrace new and emerging areas of real-time systems research.
Dr. Liu, Zheng Dong, and their fellow researchers’ papers mainly investigate the problems of ensuring predictable real-time computing in embedded systems. Two of Dr. Liu’s papers, which includes one co-authored with Zheng, will be appearing in the RTSS main track. The paper with Zheng is titled “Enabling predictable wireless data collection in severe energy harvesting environments,” and will appear in the Internet-of-Things (IoT) track. Below is a list of the three papers that were accepted into the IEEE RTSS’16.
- Zheng Dong, Yu Gu, Jiming Chen, Shaojie Tang, Tian He, and Cong Liu. Enabling predictable wireless data collection in severe energy harvesting environments, Proceedings of the 37th IEEE Real-Time Systems Symposium (RTSS), 2016.
- Zheng Dong and Cong Liu. Closing the loop for the selective conversion approach: a utilization-based test for hard real-time suspending task systems, Proceedings of the 37th IEEE Real-Time Systems Symposium (RTSS), 2016.
- Jianjia Chen, Wenhung Huang, and Cong Liu. k2Q: A quafratic-form response time and schedulability analysis framework for utilization-based analysis, Proceedings of the 37th IEEE Real-Time Systems Symposium (RTSS), 2016
Dr. Liu, an expert in real-time systems, also teaches courses in that area. These courses mainly address the fundamental concepts in real-time scheduling theory and real-time operating systems. System-oriented course projects are required to help students gain experience in building and improving real-time operating systems. “One of my recent focused topics in real-time systems was to adopt GPUs in advanced embedded systems (e.g., autonomous vehicles) in a predictably, temporally-correct, and efficient manner,” states Dr. Liu.
Two of Dr. Liu’s PhD students, Zheng Dong and Husheng Zhou, have obtained important research results that have been published in top conferences. The students have also been recipients of prestigious internships in the area. Zheng has been interning in Huawei Research-US for two consecutive years and Husheng interned at IBM research T.J. Watson center this summer.
We recently spoke with Zheng Dong, one of Dr. Liu’s PhD students and co-author of two of the accepted papers. Below is the discussion we had with him regarding his published papers, research in which he is involved, and how he came to the decision to pursue his PhD at UT Dallas under the guidance of Dr. Cong Liu.
UTDCS: I understand that you recently had two papers that you co-wrote with Dr. Cong Liu accepted into the RTSS 2016 Conference. Dr. Liu informed us that he is not aware of any graduate student who managed to get multiple first-authored papers into the RTSS conference publication. Congratulations!
If could, please provide a brief description of both papers you wrote with Dr. Liu and the other co-authors? What was the research behind the papers?
ZD: The first paper titled, “Closing the loop for the selective conversion approach: a utilization-based test for hard real-time suspending task systems” was co-written with my advisor and UT Dallas CS Professor, Dr. Cong Liu. The paper is about the real-time scheduling theory. This problem goes back 20 years when researchers considered the impact of suspension-induced delays on real-time systems. These are tasks which interact with external devices that introduce self-suspension delays. Such delays impact schedulability quite negatively if deadline misses cannot be tolerated. Since then, researchers have been contributing many different solutions to this problem using different methods. However, the results are always very pessimistic.
In 2012, my advisor Dr. Cong Liu and his PhD advisor Dr. Jim Anderson derived a schedulability test that shows that any given soft real-time suspending task system is schedulable under global earliest-deadline-first (GEDF) scheduling with only O(m) utilization loss. However, the schedulability test for hard real-time suspending task system under GEDF scheduling was still an open problem. It is this problem that has been solved by this research paper. We found that any given hard real-time suspending task system is schedulable under GEDF scheduling if the total utilization loss is at most O(K).
The second paper titled, “Enabling predictable wireless data collection in severe energy harvesting environments,” which was co-written with Drs. Cong Liu, Yu Gu, Jiming Chen, Shaojie Tang, and Tian He, is a system paper about the internet of things (IoT). We know that in micro-powered wireless sensor networks, sensor nodes need to operate under the severe harvesting conditions with their tiny energy reservoirs. However, existing low power network stacks make a slew of design choices that limit the ability to scale down to such environments. In the 11th USENIX Symposium on Networked Systems Design and Implementation (NSDI ’14), researchers addressed these issues with QuarkNet, a backscatter-based network stack that is designed to enable continuous communication even if there is only enough harvested energy to transmit a few bits at a time. However, QuarkNet only enables end-to-end communication between one reader and one sensor node. If the reader wants to communicate with multiple sensor nodes simultaneously, QuarkNet will not work. This is because the sensor nodes may interfere with each other, when they transfer data to the reader at the same time, resulting in the transformation delay being unpredictable.
In our paper, by leveraging the prioritization algorithms and the corresponding schedulability analysis learned in real-time scheduling theory, we bound the data collection delay for each individual sensor node when the reader communicates with multiple sensor nodes. Furthermore, we bound the total collection delay. Our research enhances the efficiency of the whole network and minimizes the total transformation delay by collecting data from sensor nodes in parallel.
UTDCS: How is the student-advisor interaction between you and Dr. Liu structured?
ZD: All of my papers are under the direction of Dr. Cong Liu. We regularly have individual meetings involving only Dr. Cong Liu and me. In addition, each week we have a group meeting, involving all group members. During the individual meetings, I discuss ideas on ongoing research problems, and Dr. Cong Liu provides feedback and advice from his perspective to help ensure that the research is on a right track.
During group meetings, one student typically presents their work and leads a brainstorming session within the group to help get creative ideas for solving their research problems.
UTDCS: It would be great if you could expand on the process of writing this paper (and others), your research in real-time systems, and anything else you would like readers to know.
ZD: I would like to share how we solved the schedulability problem for hard real-time suspending task systems. This story may provide inspiration and encouragement to PhD students who are encountering challenging problems in their research. I had already been working on this problem for nearly one year. There were phases when I was more focused on the schedulability problem and then there were phases when I put my effort more onto other problems; but I never stopped working. I tried many different existing techniques over that time, but somehow I wasn’t making the progress I wanted.
Then, at the beginning of 2016, I had the transforming idea of not converting the suspensions on the task schedule at task level, but instead different ones iteratively converting the suspensions at job level. I discovered that the new schedule produced by the new converting method had a key property, which was able to help us solve the problem. This was a real breakthrough.
At first, I talked to Dr. Liu about the idea in generally rough terms. He understood it quickly and encouraged me to be clear in all details. After that discussion, I spoke with Dr. Cong Liu nearly every afternoon and we worked an entire month to derive all the proofs of this paper. Still, it took us nearly 4 months of hard work until the paper was complete.
I am lucky to have Dr. Cong Liu as my advisorand he continues to give me valuable suggestions for my research. This has been a great help to me in writing many new and exciting papers.
UTDCS: On your UT Dallas website you mention that you received your M.S. degree from the University of Science and Technology of China in 2011 and your B.S. degree from Wuhan University in 2008. What made you decide to pursue your PhD at the UT Dallas CS Department? I understand your research interests comprise Real-time systems with a special focus on Real-time scheduling theory. Can you please explain this and why you decided to pursue this area of research?
ZD: The most practical reason for why I chose to pursue a PhD was so that I could use the time that it took to obtain a PhD, to focus on what I liked doing and at the same time to avoid the direct pressures of the market to pursue work with only short-term financial value. When short-term market pressures do not bind me, my imagination tends to be dramatically more creative. This may explain why I have only two papers accepted into the RTSS’16 conference. However, I like to think that there is a tiny, tiny, tiny chance that what I work on will change the world decades into the future. To me, that is pretty cool!
After I received my M.S. degree from the University of Science and Technology of China in 2011, I was invited by Dr. Yu Gu to become a research assistant at the Singapore-MIT International Design Center. During my time there, I conducted research on how to enhance efficiency of cyber-physical systems there. While working at Singapore-MIT International Design Center, I met Dr. Cong Liu for the first time via skype. Dr. Cong Liu frequently collaborated with Dr. Yu Gu on research topics. When Dr. Liu spoke to me about his research areas, I was very impressed by the real-time scheduling theory and was eager to learn more about it directly from Dr. Liu.
After I joined Dr. Cong Liu’s research group on real-time systems, I learned that there was so much going on at the UT Dallas CS Department and was intrigued by the growing department, its distinguished faculty, and the research conducted there.
The atmosphere here at the UT Dallas CS Department contributes significantly to my work. On campus there are numerous department-sponsored talks and events that bring researchers and tech companies from all over the world to UT Dallas. It is all very inspiring. All these aspects contributed to my decision to pursue my PhD at the UT Dallas CS Department and why I decided to pursue this area of research. Currently, Dr. Cong Liu is my PhD advisor.
UTDCS: What is the focus for your PhD?
ZD: As mentioned on my website, I am broadly interested in Real-time systems with a special focus on Real-Time Scheduling theory. More specifically, my PhD focus is improving processor utilization on multiprocessors for different task models by providing insightful hard or soft real-time guarantees for different real-time applications, whose preemption and migration overheads can span different ranges and whose tolerances to deadlines or delay bounds are different. The objectives can be achieved by designing new algorithms, simplifying optimal algorithms, and developing new validation tests.
UTDCS: You are a part of the real-time systems group with Dr. Liu. What are some of the group’s activities? How many other students are in the group with you?
ZD: Currently, we have four PhD students and two Masters students in our group, including myself. We perform cutting-edge research on Real-Time Computing and Real-Time Operating Systems, GPU Resource Sharing and Virtualization in the Cloud, Battery Management for Electric Vehicles and Soft Real-Time Offloading in Mobile-Cloud Systems.
We are always looking for bright and self-motivated PhD students to join our group!
ABOUT THE UT DALLAS COMPUTER SCIENCE DEPARTMENT
The UT Dallas Computer Science program is one of the largest Computer Science departments in the United States with over 2,100 bachelor’s-degree students, more than 1,000 MS master’s students, 150 PhD students, and 86 faculty members, as of Fall 2016. With The University of Texas at Dallas’ unique history of starting as a graduate institution first, the CS Department is built on a legacy of valuing innovative research and providing advanced training for software engineers and computer scientists.