The IEEE/ACM IWQoS (International Symposium on Quality of Service) was held in Florida, USA, in June 2023. IWQoS is an important international conference in the computer system community about the quality of service, and becomes an important part of the quadrennial ACM FCRC 2023 conference. Prof. Yu Hua from Huazhong University of Science and Technology, working with McGill University, published a paper titledCATS: A Computation-Aware Transaction Processing System with Proactive Unlocking, which won the Best Paper Award. IWQoS 2023 receives 264 submissions, from which one best paper is selected.
Best Paper Award Certificate in IWQoS 2023
With the rapid development of new network applications such as cloud computing, blockchain, and metaverse, the computer systems as underlying infrastructure have become more and more complex, and higher QoS requirements have been put forward. Transaction processing systems have received many attentions due to salient features, such as easy-to-use operations and atomicity guarantee. In distributed transaction processing systems, computation-related operations play an important role. However, existing transaction systems are limited by concurrent computing operations, which weakens the system scalability and efficiency, thus resulting in high latency on the critical path of concurrent control and posing new challenges to efficient transaction processing systems.
Ph.D student Bolun Zhu, advised by Prof. Yu Hua, proposes a novel transaction processing system for computational operation optimization, called CATS. Through the data dependency analysis of the access patterns and computing operations in the transaction, this work identifies that only some key operations in the transaction system need to exist on the critical path of concurrent control. Based on the above observational analysis and system verification, this work proposes a new way to identify critical operations at runtime based on the data dependencies of the program, and proactively delays the execution of non-critical operations, thereby unlocking some of the shared data in advance. In view of a variety of program errors that may be caused by delayed execution, CATS uses dynamic detection and rollback to ensure the accuracy of program execution. Evaluation results show that compared with state-of-the-art transaction processing systems, CATS can significantly improve the system performance in multiple computation-intensive and high-concurrency scenarios.
The data structure of CATS
From the establishment more than 30 years ago, IWQoS conference has improved the rapid development of many key technologies, such as cloud computing systems, audio and video network transmission, machine learning systems and operating system in terms of the quality of service, and becomes an important international conference in the quality of service for computer systems, which has extensive and far-reaching influence on the international academic and industrial communities.
Prof. Yu Hua’s research team
Professor Yu Hua's team has focused on the research of high-performance storage systems and architecture for more than ten years, and the research achievements have won the first class award in natural science of the Ministry of Education, the first class award in science and technology progress in the Hubei Province and the second class award in electronics information science and technology of the Chinese Institute of Electronics. The advised students have won the ACM China doctoral dissertation award (national winner), the outstanding doctoral dissertation award nominee in CCF and the excellent Master thesis in Hubei Province.
