IEEE/ACM IWQoS (International Symposium on Quality of Service) kicked off in Florida, the US at the end of June. IWQoS is an important international conference on the quality of service for computing systems, also an important part of ACM FCRC 2023, a quadrennial meeting held by the Association for Computing Machinery (ACM). A team led by Prof. Hua Yu from Wuhan National Laboratory for Optoelectronics, HUST and McGill University co-published a research paper at the conference titled CATS: A Computation-Aware Transaction Processing System with Proactive Unlocking, which has been the only winner of the Best Paper Award at the conference. IWQoS 2023 received a total of 264 papers and only selected one of them as the Best Paper Award winner.
As cloud computing, blockchains, metaverse and other new Internet applications develop fast, computing systems, as the underlying structure, have become more and more complicated, which comes with higher requirements for the quality of service. The transaction processing system (TPS) has become well known for being easy to use and atomic. In the distributed transaction processing (DTP) system, computing-related operations have been playing an important role. However, limited by concurrent computing, the current transaction processing system has become less efficient with respect to the scale and effectiveness. That leads to higher delay on the critical path of concurrency control and poses new challenges to the high-efficient compute-optimized transaction processing system.
Guided by Prof. Hua Yu, the HUST PhD student Zhu Bolun has developed a compute-optimized transaction processing system called CATS. In the research, the team conducted data dependence analysis for load/store and computing operations in transactions, and therefore discovered that only part of the critical operations in the transaction processing system need to be stored on the critical path of concurrency control. Based on the above-mentioned analysis and verification, the team developed a new method where critical operations are identified based on the data dependency of a running program, and non-critical operations are actively delayed so as to unlock part of the shared data early. As for potential program errors caused by the delays, CATS adopted the measures of dynamic detection and forced revert to ensure accurate execution. The testing results show that compared to the existing cutting-edge transaction processing system, CATS is capable of considerably enhancing system performance in multiple scenarios, such as dense computing and high concurrency.
Ever since its founding over 30 years ago, IWQoS has boosted the fast growth of multiple critical technologies, including cloud computing systems, audio & video transmission, machine learning systems and the quality of service for operating systems. It is an important international symposium on the quality of service for computing systems, exerting a great and extensive influence on both the international academic community and industry.
Team led by Prof. Hua Yu
The team led by Prof. Hua Yu has been dedicated to research on High Performance Storage System (HPSS), computer architecture and so on for over a decade. The team’s research results have won the First Award of Natural Sciences awarded by the Ministry of Education, the First Prize of the Hubei Science and Technology Progress Award, and the Second Prize of the CIE Electronic Information Science and Technology Award, and so on. The team’s graduate students have won awards such as the ACM China Doctoral Dissertation Award，the nomination for the CCF Outstanding Doctoral Dissertation Award, and the Hubei Provincial Outstanding Masters' Thesis Award.
Source: Wuhan National Laboratory for Optoelectronics
Edited by: Chang Wen, Peng Yumeng