Congratulations to Dr. Gabriel Parmer, Associate Professor of Computer Science, and his students, Wenyuan Shao, Bite Ye, Huachuan Wang, and Yuxin Ren, for receiving the best paper award at the 43rd IEEE Real-Time Systems Symposium (RTSS) for their paper, “Edge-RT: OS Support for Controlled Latency in the Multi-Tenant, Real-Time Edge” ! IEEE RTSS is a premier conference in the field of real-time systems and a venue for researchers and practitioners to showcase their innovative ideas regarding all aspects of real-time systems, including theory, design, analysis, implementation, evaluation, and experience.
This study is part of a CPS grant focused on making edge computing useful for the control of the physical world. Edge computing is the idea that many robots, Internet of Things (IoT) devices, and Autonomous Vehicles (AVs) either don’t have the beefy computers that they might want to do intelligent computations, or don’t have the ability to coordinate tightly together. Imagine quadcopter that must minimize weight and battery drain for an example of the former, and AVs that would benefit from coordinating with each other on things such as, “I’m braking hard now, if you’re behind me slow down now!,” as an example of the latter.
Conversely, edge computing is also the idea that we might have micro-cloud datacenters co-located with wireless-base stations that we use to connect to the internet with mobile. The grant, then, asks if we can put computations in the edge that can enable “offloading” of computations from computation-starved devices, while also providing an intelligent means of coordinating between different devices. Core challenges to this are how the software on the edge can enable clients that do not in any way trust each other to both run computations in that edge environment and how the edge software can provide replies in real-time to the devices while still packing a ton of useful computation into the edge. The latter challenge is what this paper addresses.
“Edge clouds are physically closer to the client so the client devices can expect a response time on the order of several milliseconds. During our previous work on providing system-level infrastructure for edge clouds, we found very few of the current existing edge infrastructures prove deadline-based scheduling, therefore they have no guarantee that a client device can receive the response from the edge cloud in a certain time. This really inspired us to provide end-to-end deadline aware scheduling on the edge cloud,” Shao explained.
The team’s hope for this paper was that the success of cloud computing can span down to the edge, where there are vastly fewer computational resources, while enabling the edge to work in real-time with devices that interact with the physical world. This should enable IoT devices and quadcopters to offload computations which could be cheaper, last longer on battery, and enable larger devices to coordinate. Returning to the AV example, this coordination would help increase the flow of traffic, prevent accidents, etc.
The process they followed to get this award-winning paper ready for the IEEE RTSS conference started with identifying some aspects that current systems or research cannot do well; for example, for this paper they wanted to optimize the real-time responsiveness of edge clouds. Then to solve this problem, they came up with some optimizations, such as Edge-RT, and validated their approaches by typically doing emulations or
small demos. After that came implementing these ideas and getting initial results that further prove the approach is correct, which they can then use as a guide to start writing the paper itself. In the meantime, they began doing the actual evaluation of their system and getting results, which were shown in the final paper.
“The research moves forward only due to the hardwork and brilliant innovations of the student researchers. This award was given to the paper with a student as the first author that was determined to be the best of the conference. It is a testament to the strength of the results that Wenyuan and the other students on the paper were able to engineer. When building a complex system (our system is over 100,000 lines of code), it is only through a student researcher’s persistence, dedication, passion, and technical impeccability that such innovative steps can be made," Professor Parmer stated.
Shao’s words on what this award means to him were, “It indeed means a lot to me. It is really an honor to be recognized in this community. Also, since I’ve spent a ton of time and effort on this project, I view this award as a praise of all the effort from my colleagues and I. Finally winning this award encourages me to devote more effort to our future work.”