Upper Index FTL Strategy

22 Jun

In order to overcome the unique physical characteristics of flash memory to index design challenges, a lot of work to study how to improve read and write performance of flash memory. These efforts can be divided into block mapping technology and pure flash index two aspects.

Block mapping technology primarily from FTL technology. FTL technology through the physical address of the memory mapped to the logical address, providing the same IO interface and disk, cover up flash offsite updates. Much work is committed to improving the performance of FTL. These efforts mainly from the following two aspects improved: Group commit strategy and adjust to fit the upper index FTL strategy. Group submitted strategy will randomly write in memory buffer at the appropriate time will be written to the flash memory random write batch. Typical in this regard the work BFTL, NFTL and FAST. BFTL achieved in a random write of a page when the data is written to the flash memory; NFTL BFTL will be the basis in the memory of the same data to merge multiple operations; FAST data is stored in the cold heat in different places, by reducing cold Data The movement thereby improving the performance of writing. Other index structure on top of regulation in FTL index structure to take advantage of existing FTL, to improve performance. FlashDB [10] data into write-intensive and intensive data read and write intensive data storage mode to take the log, read-intensive data taken disk storage mode. Log storage mode takes full advantage of the characteristics of FTL read and write speeds to achieve better balance. Has put forward for the log deletion and recovery methods.

Block mapping technique is mainly due to its lack of performance simulate the flash disk, leading to its high-speed access feature is not fully exploited. Therefore, an index of pure flash memory directly through the physical address to access the data. Pure Flash index generally take the log to record updates,

Thereby reducing the cost of writing. JFFS3 in memory maintains a log tree to delay and reduce the number of writes. Taken offsite data in the flash memory update way to reduce the cost of in situ update. IPL then take the data and its associated logs are stored in the same block. Also based on embedded devices LGeDBMS [16] also adopted a log-based approach to recording data updates, so DBMS managed directly bypassing the file system on the flash data to improve database performance.

Overall, the block map and pure flash indexing techniques can not play a high-speed flash memory access performance, this paper presents a non-log-flash index structure to solve the problems in flash index.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: