"Self-powered Internet-of-Things Nonvolatile Processor and System Exploration and Optimization"

Kaisheng Ma, Ph.D: Penn State U.

March 21st (Wednesday), 11:00am
Harold Frank Hall (HFH), Rm. 4164 (ECE Conf. Rm.)

Energy harvesting has been widely investigated as a promising method of providing power for ultra-low-power applications. Such energy sources include solar energy, radio-frequency (RF) radiation, piezoelectric effect, thermal gradients, etc. However, the power supplied by these sources is highly unreliable and dependent upon ambient environment factors. Hence, it is necessary to develop specialized systems that are tolerant to this power variation, and also capable of making forward progress on the computation tasks.

In this talk, I will first do architectural explore of the design space for a nonvolatile processor with different architectures, different input power sources, and policies for maximizing forward progress. It is presented that different complexity levels of nonvolatile microarchitectures provide best fit for different power sources, and even different trails within same power source.

To further overcome this problem, I further propose various techniques including frequency scaling and resource allocation to dynamically adjust the microarchitecture to achieve the maximum forward progress. Noticing that such nodes usually perform similar operations across each new input record, which provides opportunities for mining the potential information in buffered historical data, at potentially lower effort, while processing new data rather than abandoning old inputs due to limited computational energy. This approach is proposed as incidental computing, and synergies between this approach and approximation techniques is explored. Last but not least, I take fog computing in Wireless Sensor Networks (WSN) as one of the system level examples to perform optimization from programing, intra-chain and inter-chain level, and show how nonvolatility features including nonvolatile processors and nonvolatile RF can benefit the system, and how other optimizations like load balance under
unstable power, as well as increasing nodes density for quality of service can be applied into the fog computing system.

About Kaisheng Ma:

Photo of Kaisheng Ma Kaisheng Ma is now a Ph.D. in Department of Computer Science and Engineering, The Pennsylvania State University. His research focuses on computer architecture, especially on IoT Fog Computing architecture exploration and optimization. For the first time, Dr. Ma explores the energy harvesting nonvolatile processor design, including tradeoffs between re-execution and backup penalty, architectural complexity vs. performance, etc. Two outstanding findings: a, optimizing for low power is not a good fit for energy harvesting. b, different application scenarios have different best architectural selections. Because of Dr. Ma’s research, for the first time, NVP architectural tradeoffs was introduced into the researchers within architectural level. The related work was published on HPCA 2015 Best Paper and and IEEE MICRO Top Picks 2016. In Micro 2017, for the first time, incidental approximate computing concept is proposed for energy harvesting scenarios, for a phenomenon is observed: a partial low quality outputs can sometime be more useful and urgent than delayed best quality outputs. System level optimizations for energy harvesting scenarios, based on nonvolatility is proposed by Dr. Ma, and the related paper will be publishing in ASPLOS 2018. In the past 5 years, Dr. Ma has published 35 papers (half first author), and 362 google citations (Feb 2018), and has several patents in US. As first author, Dr. Ma has won many awards, including: 2015 HPCA Best Paper Award, 2016 IEEE MICRO Top Picks, 2017 ASP-DAC Best Paper Award. Dr. Ma has many honors, including 2016 Penn State CSE Department Best Graduate Research Award (Among ~170 Ph.D. students), 2016 Cover Feature of NSF ASSIST Engineering Research Center Newsletter (Among 40 graduate students across four participating universities.), 2011 Yang Fuqing & Wang Yangyuan Academician Scholarship (1/126, Peking University.). His research interests include Non-volatile processor architecture and Neural Networks Accelerator Design.

Hosted by: UCSB CE Program