Cross-Layer Optimizations for Fast and Secure Storage Systems
In this Big Data era, modern data-intensive applications are generating unprecedented amounts of data, while demanding high performance and high security of computing systems. Memory and storage systems play a key role in modern computing systems to meet such requirements, as they directly deal with the large amounts of data. Unfortunately, there are increasing difficulties in scaling memory and storage devices (e.g., DRAM and NAND flash memory), introducing significant performance and reliability issues. To bridge this gap between applications and memory/storage devices, we need more intelligent system software and computer architectures that are co-designed across the system hierarchy with a deep understanding of both high-level application requirements and low-level device characteristics.
This talk presents two recent works that propose cross-layer optimization techniques to improve the performance and security of modern NAND flash-based SSDs. The first work tackles the long and non-deterministic read latency due to read-retry, an essential reliability-enhancement technique in modern NAND flash-based SSDs. The work introduces Pipelined and Adaptive Read-Retry (P&AR2) that reduces the read-retry latency by leveraging the read-retry characteristics of modern NAND flash devices and the reliability margin existing in modern SSDs. The second work tackles the data-remanence problem in NAND flash-based SSDs where deleted data can remain in the SSD for an indefinite time. The work introduces an access control-based data sanitization technique, called Evanesco, which enables a NAND flash chip to be aware of each data’s validity and avoid data transfer of invalid data.
Jisung Park is an Assistant Professor in the Department of Computer Science and Engineering at POSTECH where he joined in September 2022. Prior to joining POSTECH, he was a Postdoctoral Research Associate and Lecturer in the Department of Information Technology and Electrical Engineering at ETH Zürich. He received his Ph.D. degree in Electrical Engineering and Computer Science and B.E. degree in Computer Science and Engineering from Seoul National University. His research interests span a wide range of the computer systems fields, including computer architecture, system software, and embedded systems, with a special focus on memory and storage systems, system security, and hardware-software interaction. His work has been published in many prestigious computer systems conferences, including ISCA, MICRO, ASPLOS, HPCA, USENIX FAST, and DAC, and has been recognized with Samsung Humantech Paper Award (Bronze Prize in 2016) and Best Paper Award Nomination at DAC 2016.