WOSAR 2024: 16TH INTERNATIONAL WORKSHOP ON SOFTWARE AGING AND REJUVENATION
co-located with ISSRE 2024, October 28-31, Tsukuba, Japan
co-located with ISSRE 2024, October 28-31, Tsukuba, Japan
Keynote 1
Multi-granularity Software Rejuvenation for Dependability Control of Service Chain
Xiaolin Chang
Abstract
Service Chain refers to a group of service components working collaboratively to handle one or more tasks. However, software aging can lead to performance degradation of service components, even failure, which ultimately degrade the performance and dependability of services chain. This talk features the semi-Markov process–based approach for analyzing the service-chain dependability under multi-granularity software rejuvenation in the dynamic and heterogeneous system. We will discuss the derivation of transient and steady-state metric formulas for the n-sized service chain, along with their application in dependability bottleneck identification and dependability optimization. Finally, we will outline current challenges and propose future research directions.
Biography
Xiaolin Chang is a Professor in the Department of Information Security at Beijing Jiaotong University, China. She received her Ph.D. degree in Computer Sciences from Hong Kong University of Science and Technology. In addition to more than 15 years’ academia research experience, she has about ten-year industrial research and development experience. Her current research interests include secure and trustworthy intelligent computing with application to Blockchain, cloud data center and vehicular networks. She has (co-)authored over 180 papers in peer-reviewed international journals and conferences. She has received several awards, including Science and Technology Progress Award of Chinese Academy of Sciences (The First and Second Class in 1997 and 1999, respectively), National Science and Technology Progress Award of China (The Second Class in 2000), Ministry of Education Science and Technology Progress Award of China (The First Class in 2010), Huaxia Construction Science and Technology Award (The Second Class in 2021), and China Invention Association Invention and Entrepreneurship Innovation Award (The Second Class in 2023 and 2024), the Best Paper Award at IEEE Globecom 2018 and ICSOC 2022. She serves on editorial boards of IEEE IoTJ and two Chinese journals.
Keynote 2
Beyond Reboot-based Recovery: Making System Software Resilient with Software-surgery
Hiroshi Yamada
Abstract
Reboot-based recovery is a simple but powerful way to improve the availability of computer systems. Periodic reboots of the target software, a.k.a. software rejuvenation, reclaim stale or leaked resources such as memory and descriptor leak to proactively prevent failures such as crashes and hangs. Reboot-based recovery faces a challenge to apply it to modern in-memory applications like machine learning, in-memory databases, and graph processing. Rebooting in-memory applications causes a non-trivial performance penalty compared to stateless applications such as web servers. The in-memory applications allocate hundreds to thousands of gigabytes of memory and manage numerous running states in their own memory regions. Since their reboots involve the loss of running states preserved in memory regions, restoring the running states from storage, replica, or recalculation is required.
Our research group is actively exploring software-surgery-based recovery that extends the concept of reboot-based recovery. Unlike conventional reboot-based recovery that reboots the entire software, software-surgery prunes damaged memory objects and reconstructs the internal structures by using only the undamaged ones at runtime. This approach promises ultra fine-grained reboot-based recovery with almost zero downtime and no time-consuming state restoration for recovery. This keynote presents an overview of the software-surgery approach and its two case studies. First is VampOS, a unikernel that performs efficient reboot-based recovery of the unikernel layer. Second is Memcached-ss, an in-memory key-value store that actively reconstructs its internals to recover from ECC-uncorrectable memory errors without reboot-based recovery.
Biography
Hiroshi Yamada is an Associate Professor at the University of Agriculture and Technology, heading the Advanced Systems Group (ASG). His research interests include operating systems, virtualization, dependable systems, and cloud computing. He is a member of ACM, IEEE, and USENIX.