DocumentCode :
3540932
Title :
On the distribution of valid pages with greedy garbage collection for NAND flash
Author :
Peleato, Borja ; Agarwal, Rajiv ; Cioffi, John
Author_Institution :
Electr. Eng. Dept., Stanford Univ., Stanford, CA, USA
fYear :
2012
fDate :
5-8 Aug. 2012
Firstpage :
500
Lastpage :
503
Abstract :
Flash memory uses relocate-on-write, also called out-of-place write for performance reasons. Relocate-on-write requires a garbage-collection process, which results in additional write operations referred to as write-amplification. The reclaiming policy that selects the blocks to garbage-collect is greedy if it is based only on the amount of free space to be gained-Write amplification is a critical factor limiting the write performance and endurance in solid-state drives and the greedy garbage collection policy minimizes it. In this paper, we derive an expression for the steady-state distribution of valid-pages using a novel probabilistic model for flash-based SSDs for a greedy reclaiming policy. This distribution lends itself directly to evaluation of write-amplification and gives insight into designing newer garbage collection policies, which would further reduce write-amplification. Numerical evaluation of the derived expression matches the distribution obtained via Monte-Carlo simulation for several different drive configurations, thereby validating the derivation.
Keywords :
Monte Carlo methods; flash memories; greedy algorithms; paged storage; performance evaluation; Monte Carlo simulation; NAND flash; block selection; flash memory; flash-based SSD; free space; greedy garbage collection; greedy reclaiming policy; numerical evaluation; out-of-place write; probabilistic model; relocate-on-write; solid-state drives; steady-state distribution; valid page distribution; write amplification; write performance; Ash; Equations; Flash memory; Mathematical model; Monte Carlo methods; Numerical models; Steady-state;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Statistical Signal Processing Workshop (SSP), 2012 IEEE
Conference_Location :
Ann Arbor, MI
ISSN :
pending
Print_ISBN :
978-1-4673-0182-4
Electronic_ISBN :
pending
Type :
conf
DOI :
10.1109/SSP.2012.6319743
Filename :
6319743
Link To Document :
بازگشت