As the popularity of flash memory increases, solid state drive (SSD) based on flash memory is widely used in the enterprise computing environment, owing to its shock-resistance, non-volatile, low energy consumption, and high I/O speed. Meanwhile, SSD has the characteristics of not-in-place update and asymmetric I/O latency among read, write, and erase operation as write/erase operation are much slower than read operation. Hence, the buffer replacement algorithm needs to properly reflect the asymmetry. This paper proposes a particle swarm optimization (PSO)-based buffer management algorithm to accurately predict each logical page in the buffer as hot or cold. The prediction is a key criterion of page-level FTL scheme. The pages predicted as hot are maintained in the buffer to maximize the hit ratio of SSD buffer. Simulation results show that the proposed scheme outperforms the existing schemes in terms of read and write hit count and buffer hit ratio.
[1]
Riccardo Poli,et al.
Particle swarm optimization
,
1995,
Swarm Intelligence.
[2]
Anand Sivasubramaniam,et al.
Leveraging Value Locality in Optimizing NAND Flash-based SSDs
,
2011,
FAST.
[3]
Lihua Yue,et al.
Flash memory management based on predicted data expiry-time in embedded real-time systems
,
2008,
SAC '08.
[4]
Joonwon Lee,et al.
CFLRU: a replacement algorithm for flash memory
,
2006,
CASES '06.
[5]
Vivek S. Pai,et al.
SSDAlloc: Hybrid SSD/RAM Memory Management Made Easy
,
2011,
NSDI.