Data Compression Capabilities with FTL

9 Sep

Flash memory is widely used in recent years, a semiconductor memory, an important characteristic of flash memory is to modify the data can not be written as the covering (over-write), i.e., when the need to modify the page (page) data, not like disk as directly make changes in place, and the page needs to be erased and re-written. delayed due erase larger, generally writes new data to another blank page.

  • Flash memory is erased for a limited number of features, based on LZ77 compression algorithm uses a strategy based on a single page compression address mapping scheme, design and implementation of a data compression capabilities with Flash Translation Layer-FTL.
  • In the data read / write memory before the data compression / decompression, can reduce the amount of data actually written to flash memory, while reducing the use of cache read and write latency.
  • Experimental results show that the actual write FTL and can effectively reduce the amount of data read, write bandwidth increases dramatically, flash SSDs have greatly improved read and write latency, available storage space than the actual capacity increase 40%.
Advertisements

NAND Flash Memory in Parallel Storage

30 Aug

With the continuous development of integrated circuit technology, the microprocessor continuously improve performance, storage system performance gradually become a constraint entire computer system performance bottlenecks. Traditional disk device due to its larger I / O access latency, has gradually unable to meet future application requirements. Flash memory read and write speed, non-volatile and other features, has been widespread concern, the current NAND flash-based SSD has been widely applied.

  The main research topics of NAND flash memory storage system and its key performance optimization techniques.

  • First, the design of parallel NAND flash-based storage system, its main features are: multiple levels of parallelism and the proposed pipeline techniques to improve memory access performance; taken a high-speed PCI Express 2.0 interface to connect to a host device; taken as a large-capacity DRAM volatile write buffer, and in the system to complete the DDR memory controller design.
  •   Secondly, given the characteristics erase before write lower write performance of NAND flash memory, NAND flash memory in the parallel storage system presents a log-based flash translation layer block offset priority strategy, efforts to reduce the system to write to the flash or erase operation.
  • Experimental results show that the strategy can not only reduce the flash erase operation, and can reduce the overall implementation of memory garbage collection time and the average latency and improve the lifetime of the flash, and improve the access performance of the flash memory.
  •   Finally, in parallel NAND flash memory storage systems add cache memory to improve performance is an effective solution to this problem is also proposed based on block-level LRU cache replacement strategy and shift priorities of integrated management strategies, with certain flash-oriented caching strategies were compared, can further improve the memory access performance.

FTL BAST Policy and Strategy

1 Aug

As users demand access to data, the traditional structure of the system and the external storage disk I / O delay limits the storage system performance. In order to meet the needs of high-performance mass data storage, new storage devices came into being. NAND flash-based solid state disk with its non-volatile, fast access, high reliability and low power consumption and many other advantages, can better alleviate the I / O performance bottlenecks, and gradually become a research hotspot. But before writing erasure characteristics and limited erase cycles a serious impact on the performance and reliability of SSDs.

  • In this paper, flash-based SSDs technical characteristics and the internal composition of the flash-based SSDs some of the key technology improvements. The research work mainly includes the following aspects:
  • Studied the flash-based SSDs basic principles of the structure and key technologies;
  • Disksim simulator designed and implemented based on the SCSI interface and flash memory interface;
  • Studied the flash-based SSD Cache works, focused on the Cache replacement policy. By analyzing CFLRU defect replacement policy proposed CFLRU / AF-E strategy, frequency of visits and page by wear leveling into account, effectively improve the Cache hit rate and reduce the number of write-back, good wear leveling obtained better degrees. Then design experiments from Cache hit rate, the total number of write-back, with a total running time and wear-leveling degree four angles CFLRU / AF-E strategy and tactics CFLRU comparative analysis;
  • Brought before the erase flash memory rewriting for wear problems, the first analysis of the Dual-Pool advantages and disadvantages of wear leveling algorithm proposed Advanced Dual-Pool algorithm, improved its position in the lack of garbage collection, wear leveling does not affect the treatment effect the premise of improving the efficiency of garbage collection. Through simulation experiments proved Advanced Dual-Pool algorithm is better than in the Dual-Pool algorithm processing performance in wear leveling similar, but garbage collection efficiency has been greatly improved;
  • Designed a flash translation layer strategy. FAST from the analysis of strategies and tactics BAST defects, combining their strengths, proposed a hybrid FTL strategy, a combination of hot and cold properties of the data block consideration solves FAST log block is exhausted in the idle poor performance, and effectively improves write performance. Through simulation experiments proved that hybrid FTL policy strategies with respect BAST and FAST strategy has greatly improved overall performance, and can effectively reduce FTL strategy caused by the number of erase operations.

In summary, the article from the SSD Cache technology, wear leveling and flash translation layer strategy three aspects of research, improved CFLRU / AF-E strategy to improve the Cache hit rate and reduce the number of write-back, improving write performance; proposed algorithm Advanced Dual-Pool obtain a higher recovery efficiency; the hybrid strategy combines FTL BAST policy and FAST strategy has the advantage, to effectively improve the write performance.

FTL Strategy Key Functions

26 Jul

As a result of flash memory chips as the storage medium, NAND Flash SSD flash memory inherited the unique hardware features. Flash memory chip read / write (program) as the basic unit of operation in pages, erasing a block as the basic unit, usually composed of block consists of several pages, written by the page must be erased before writing, as well as for the same block the erase cycles is limited, beyond the prescribed number of erase some blocks will advance damaged.

  • Flash memory is a serious impact on these hardware features SSD write speed, longevity and reliability main performance. SSD flash SSDs wear leveling has been a hot research topic.
  • According to the block erase flash memory loss is determined as the balanced performance indicators, on this basis, the use of FTL from the dynamic wear leveling and static wear leveling strategies the two aspects of flash memory loss strategy equilibrium of the system analysis and design.
  • SSD flash memory loss given equilibrium strategies take into account the dynamic wear leveling and static wear-leveling, with a minimum erase count block allocation strategy, garbage collection strategy and static wear-leveling strategies to make flash physical block erase all times consistent trend.
  • In addition, for flash-based storage characteristics to achieve a solid-state drive storage management. SSD memory management system to the upper shield can flash the internal characteristics, to achieve wear leveling strategies and standard block device access interface.

In the analysis of SSD storage management architecture based on data from the address mapping mechanism, bad block management and data storage security, fast startup and brownout recovery and other aspects of SSD storage management implementation techniques are discussed. Simulation results show that this strategy FTL can make all the blocks in the flash memory erase counts uniform, and significantly increase the life of the SSD.

Flash Translation Layer(FTL) and Underlying Structure

23 Jul

In recent years, due to its non-volatile flash memory, low power, access to high-speed, small size and other characteristics has been widely used, but it is neither a character device, nor is it a block device, its unique physical characteristics make it in the traditional file system application of certain obstacles.

Its characteristics are also included: a limited number of block erase and erase rewritable, which will affect the performance and service life of flash memory. How to provide a proper flash file system to improve memory performance and extend the life of flash memory, as many scholars focus of the study.

In this paper, the flash file system for a comprehensive study of the evolution of history, summed up the various underlying flash file system mechanism and performance analysis.

  • First, the physical characteristics of the flash are introduced;
  • Then study the FTL Flash Translation Layer and various flash file system and its underlying structure, summed up the flash file system commonly faced three major issues: garbage collection, wear leveling and bad block management;
  • Then proposed based on a novel high frequency data interchange wear leveling algorithm OEBS, through the experimental data show that effectively balances the erase distribution, improving the flash file system to extend the life of degree;
  • Flash file system on the forefront of the underlying mechanisms have been studied and made a quantitative performance analysis; concludes in its entirety on the flash file system performance of the underlying mechanisms and research summary and outlook.

FTL Block Erase & Data Storage

18 Jul

 

Flash memory (flash memory) as a With the rapid development of the microelectronics industry and the birth of new semiconductor memory device, with its unique non-volatile, high density, low cost, gradually replacing the ROM / EPROM / EEPROM, in recent years has been rapid development. Especially the Serial Flash (Serial Flash), because it has less data pin, a simple peripheral circuit constructed so that it is more commonly used by people.

 

However, due to its memory storage architecture reason, when the flash write data from 1 to 0, to 0, it can not be changed by writing a 1. When the flash memory erasing data from 0 to 1, can not be erased only by a unit. When erasing and rewriting its data, only the smallest unit of a block operation, including flash erase block erase and chip erase.

 

FTL a block erase is to erase the contents of the block are changed to 1, the chip is to erase the entire contents into a flash. Usually a flash memory chip is divided into a plurality of block erase, during the flash memory in units of erase blocks. And it has 100 000 -100 Million endurance life. These characteristics of flash memory devices are different from the other.

 

Therefore, the subject of these characteristics for the flash memory, FTL on flash (serial) data storage application interface in the design and management of data files made ​​corresponding research and thinking, mainly on the client / manufacturer in the application of serial Flash in the process of starting the practical problems encountered in the specific, such as data loss, misuse instruction, CRC checksum errors, design and preparation of a company based on NI LabWindows / CVI platform for M25 series serial Flash diagnostic data analysis tools to facilitate efficient and easy to use client. And the test proved that the tool is simple and practical.

 

FTL–Flash Memory Technology

16 Jul

As in recent years, flash memory (Flash Memory) memory technology, the rapid development of the capacity of flash memory devices, equipment stability, reduce energy consumption, improve read and write speeds, etc. have made a breakthrough, particularly as a fixed storage devices NAND flash memory is widely applied in many areas and gradually replace the disk, and thus need to effectively organize, store, manage and use the data in the flash memory.

  • Current data management technology mainstream database technology. Therefore, it is necessary to examine how the flash memory database system to store and manage the data.
  • Since flash memory has a limited number of erase, erase-before-write characteristics, will bring limited lifetime defects. Therefore, the management of flash memory compared to the other will be very different.
  • Current database system for embedded systems is mainly to solve personalized data storage and handling problems, they are embedded a variety of electronic devices and mobile devices, FTL is characterized by small, spiritual, easy, it is usually associated with the operating system and application-specific integrated in together, but the current system is not designed for embedded database storage media such as flash special studies and optimization.
  • Now flash-based storage management technology is mainly concentrated in the file system applications, the most important technologies including flash memory and flash translation layer log-style management techniques, they have some FTL management flash memory database management system worthy place to learn .

 Flash memory database management system needs to garbage collection and cleaning mechanism to reclaim dead space for future use, but also to be able to do wear leveling to make some quick flash life due to wear and tear will not be broken. However, the low cost and wear leveling garbage collection, it was a difficult co-existence of the goal. At the same time, wear leveling is divided into dynamic and static wear leveling wear leveling, requires different mechanisms. This main task is to present the basic memory management and garbage collection policy for flash memory wear leveling strategies and use of the database for analysis and improvement. Garbage collection algorithm added to the different types of storage units management capabilities, considering the age of the data, so that garbage collection to improve efficiency while writing data to do wear leveling, and designed specifically for the static data static wear leveling algorithm, so as to achieve more effective wear leveling. Finally, experiments show that the improved design of garbage collection and wear leveling strategies have a good memory to improve system performance and extend the life of the effect.