Haruki Imai, Tung D. Le, et al.
Big Data 2020
This paper presents analyses of various configurations of error correction codes for the purpose of reducing the parity area for quasi-nonvolatile data retention by DRAMs. By combining long and short error correction codes, we show that the parity area can be reduced to less than 1% of the total memory size, yet the system can offer comparable reliability and adaptability as an earlier design that requires 12.5% parity area. We also claim that even without using any area for parity data, the adaptive control of the DRAM refresh rate can reduce the total risk of data loss. Finally, we discuss an efficient decoder design for long RS codes. © 2000 IEEE.
Haruki Imai, Tung D. Le, et al.
Big Data 2020
Atsuya Okazaki, Yasunao Katayama, et al.
CF 2011
Yuki Ito, Haruki Imai, et al.
PPoPP 2019
Yasunao Katayama
GLOBECOM 2019