[Academics] [Seminar] Leveraging Approximate Data and Sentinel Cells for Fast Read on NAND Flash (5/22)
- College of Software and Engineering
- Hit432
- 2019-05-09
[Seminar]
Date: 11AM 22 May 2019
Venue: Semiconductor Building 400112
Title: Leveraging Approximate Data and Sentinel Cells for Fast Read on NAND Flash(Prof. Chun Jason Xue)
Abstract: With the shrinking technology size and increasing bits per cell, NAND flash is experiencing increasing number of errors,
which requires longer latency to correctly read the data.
On one hand, when using error correction codes (ECC) with strong error correction capability,
like low-density parity-check (LDPC) code, multiple reads are required to obtain soft information for data with high raw bit error rate (RBER).
This work proposes to address this issue with the assistance of approximate data, since there are ample amount of approximate
data available in flash storage. A novel data organization is proposed to fortify the reliability of regular data by leaving
approximate data unprotected. We also present a new data allocation strategy and modified garbage collection scheme to
complete the design.
On the other hand, read retry is applied to reduce RBER seen by ECC. A read failure followed by a read retry happens when
errors exceed ECC capacity. Finding the right read voltage with the smallest number of read failures is the key to read performance.
We propose to reserve a small set of cells as sentinels, from which the optimal voltage can be inferred, as drifting
caused errors exhibit strong locality.
Bio: Prof. Chun Jason Xue received the B.S. degree in computer science and engineering from the
University of Texas at Arlington, Arlington, TX, USA, in 1997, and the
M.S. and Ph.D. degrees in computer science from the University of Texas
at Dallas, Richardson, TX, USA, in 2002
and 2007, respectively. He is currently an Associate Professor with the
Department of Computer Science, City University of Hong Kong, Hong
Kong. His current research interests include memory and storage
optimizations.