315 results about "Write amplification" patented technology

Priority-based garbage collection utilizes attributes of data stored in the non-volatile memory array in order to improve efficiency of garbage collection and of the overall data storage system. A set of low priority data can be selectively evicted from a non-volatile memory array. This can, for example, reduce write amplification associated with garbage collection. Another set of low priority data can be regrouped or consolidated in a different region of the non-volatile memory array. In addition, flushing of data can be performed in order to enhance or optimize garbage collection. Performance and endurance can thereby be improved.

Apparatus and methods provide relatively low uncorrectable bit error rates, low write amplification, long life, fast and efficient retrieval, and efficient storage density such that a solid-state drive (SSD) can be implemented using relatively inexpensive MLC Flash for an enterprise storage application. A page is associated with a set of primary ECC codewords, and a page stripe is associated with a set of secondary codewords and primary over secondary parity (PoSP) ECC codewords. Two or more page stripes can form a page grid, wherein the page grid is associated with a group of tertiary ECC codewords, wherein the last page stripe of the page grid has a reduced payload capacity.

Apparatus and methods provide relatively low uncorrectable bit error rates, low write amplification, long life, fast and efficient retrieval, and efficient storage density such that a solid-state drive (SSD) can be implemented using relatively inexpensive MLC Flash for an enterprise storage application. Predefined gears correspond to different predefined ECC schemes. Based on an observed bit error rate, a gear from a set of predefined gears is selected for use for a particular region of memory. Each gear of the set of predefined gears includes a lower-latency ECC decode option and one or more higher-latency ECC decode options.

Apparatus and methods provide relatively low uncorrectable bit error rates, low write amplification, long life, fast and efficient retrieval, and efficient storage density such that a solid-state drive (SSD) can be reliably implemented using various types of memory cells, including relatively inexpensive multi-level cell flash. One embodiment intelligently coordinates remapping of bad blocks with error correction code control, which eliminates the tables used to avoid bad blocks.

An apparatus having a memory circuit and a manager is disclosed. The memory circuit generally has (i) one or more Flash memories and (ii) a memory space that spans a plurality of memory addresses. The manager may be configured to (i) receive data items in a random order from one or more applications, (ii) write the data items in an active one of a plurality of regions in a memory circuit and (iii) mark the memory addresses in the active region that store the data items as used. Each data item generally has a respective host address. The applications may be executed in one or more computers. The memory addresses in the active region may be accessed in a sequential order while writing the data items to minimize a write amplification. The random order is generally preserved between the data items while writing in the active region.

Methods, systems, and computer programs are presented for storing data in a solid state drive (SSD). One method includes an operation for detecting a plurality of streams writing to the SSD, each stream writing in sectors, a page including a plurality of sectors and a block including a plurality of pages. A write operation includes writing at least one complete page, and an erase operation includes erasing at least one complete block. The method further includes operations for allocating a write buffer for each stream in RAM memory, and for storing each received sector of a stream in the corresponding write buffer. When a write buffer stores enough sectors to fill a page, content of the write buffer is written to a page in flash memory such that the page is filled. Further, the write buffer is freed after writing the content of the write buffer to the flash memory.

Apparatus and methods provide relatively low uncorrectable bit error rates, low write amplification, long life, fast and efficient retrieval, and efficient storage density such that a solid-state drive (SSD) can be implemented using relatively inexpensive MLC Flash for an enterprise storage application.

Disclosed embodiments are directed to systems and methods for dynamic overprovisioning for data storage systems. In one embodiment, a data storage system can reserve a portion of memory, such as non-volatile solid-state memory, for overprovisioning. Depending on various overprovisioning factors, recovered storage space due to compressing user data can be allocated for storing user data and/or overprovisioning. Utilizing the disclosed dynamic overprovisioning systems and methods can result is more efficient utilization of cache memory, reduction of write amplification, increase in a cache hit rate, and the like. Improved data storage system performance and increased endurance and longevity can thereby be attained.

Apparatus and methods provide relatively low uncorrectable bit error rates, low write amplification, long life, fast and efficient retrieval, and efficient storage density such that a solid-state drive (SSD) can be reliably implemented using various types of memory cells, including relatively inexpensive multi-level cell flash. One embodiment intelligently coordinates remapping of bad blocks with error correction code control, which eliminates the tables used to avoid bad blocks.

The invention provides a data storage method and a data storage device. The method comprises the following steps of: obtaining data to be stored into a solid-state hard disk from a memory; caching the data into a magnetic hard disk in a data structure being a log structure merging tree; executing a merging operation on the data cached in the magnetic hard disk; and when the volume of the data cached in the magnetic hard disk reaches the preset volume, storing the merged data into the solid-state hard disk. By using the method or the device provided by the invention, the write amplification problem generated during the updating of small-particle-size data can be solved; and the service life of the solid-state hard disk is prolonged.

A data storage system is disclosed that utilizes garbage collection and hybrid self-mapping for managing system data. In one embodiment, a system data region having an amount of valid system data that is below a threshold is freed. Write amplification associated with managing and storing system data can be reduced at least in part because only valid system data can be copied during garbage collection of the selected region. Mapping information associating system data with physical locations in non-volatile storage where system data is stored can be generated, which can reduce system data reconstruction time during start-up of the data storage system. Increase in efficiency and reduction in startup and initialization time can be attained.

In one embodiment, a flash-optimized, log-structured layer of a file system of a storage input/output (I/O) stack executes on one or more nodes of a cluster. The log-structured layer of the file system provides sequential storage of data and metadata on solid state drives (SSDs) to reduce write amplification, while leveraging variable compression and variable length data features of the storage I/O stack. The data may be organized as an arbitrary number of variable-length extents of one or more host-visible logical units (LUNs). The metadata may include mappings from host-visible logical block address ranges of a LUN to extent keys, as well as mappings of the extent keys to SSD storage locations of the extents. The storage location of an extent on SSD is effectively “virtualized” by its mapped extent key such that relocation of the extent on SSD does not require update to volume layer metadata.

Apparatus and methods provide relatively low uncorrectable bit error rates, low write amplification, long life, fast and efficient retrieval, and efficient storage density such that a solid-state drive (SSD) can be implemented using relatively inexpensive MLC Flash for an enterprise storage application.

Disclosed is an RAID5 (redundant array of independent disk 5) write IO optimization processing method. The method includes following steps: 1), storage system creation; 2), data reading and writing operation; and 3), failure data block recovery and has the advantage of capability of effectively lowering write amplification effect of RAID level disk arrays. For conventional magnetic disks, sequential write performance is much higher than random write performance. The address mapping method enables random write operation to be converted into sequential write operation, write performance of the magnetic disk is effectively brought into play, IO operation needed for computation checking is lowered, and response capability of an RAID system is evidently improved, and power consumption of the system is lowered. In addition, an address mapping module can flexibly choose arrangement positions and the number of checking blocks, so that the RAID5 write IO optimization processing method is not limited to be applicable to disk arrays of RAID5 level and can be popularized for realization of RAID4 and RAID6 levels.

Apparatus and methods provide relatively low uncorrectable bit error rates, low write amplification, long life, fast and efficient retrieval, and efficient storage density such that a solid-state drive (SSD) can be implemented using relatively inexpensive MLC Flash for an enterprise storage application.

The invention discloses a mapping table management method for a solid state disk and the solid state disk. The method is characterized in that an SCM is added into the solid state disk, a hard disk controller regards the SCM as a solid state disk data cache, and a mapping table used for recording a mapping relation between a specific physical address of an NAND and a logic address accessed by a host is created and stored in the SCM; and when the host writes host data into the solid state disk, the data is written into the SCM first to be cached, and the mapping table is updated synchronously after the data cached in the SCM is actually updated to the NAND or the mapping table is updated synchronously when the data is written into the SCM to be cached. By storing the mapping table in the SCM, the mapping table no longer needs to be stored in the NAND, occupation of space of the NAND is avoided, and therefore the problem of writing amplification caused by frequent updating of the mapping table is avoided. Meanwhile, the operation of loading an initial mapping table from the NAND after startup is omitted, and startup response delay is greatly shortened.

Apparatus and methods provide relatively low uncorrectable bit error rates, low write amplification, long life, fast and efficient retrieval, and efficient storage density such that a solid-state drive (SSD) can be reliably implemented using various types of memory cells, including relatively inexpensive multi-level cell flash. One embodiment intelligently coordinates remapping of bad blocks with error correction code control, which eliminates the tables used to avoid bad blocks.

The invention discloses a method and device for storing data. The method specifically comprises the steps of obtaining key value pair data to be stored; storing the key value pair data in pre-write log files of a disk; generating an index entry for the key value pair data in a memorizer according to the storing position of the key value data in the disk and keys in the key value pair data so as to execute preset operation on the key value pair data through the index entry, and writing the key value pair data reserved in a pre-write log file which satisfies a preset condition into the disk when the pre-write log file which satisfies a merging settling condition is detected, wherein the storing position comprises at least the file name of the pre-write log file and offset of the key value pair data in the pre-write log file and the initial position of the file. According to the method and device for storing the data, the problem of write amplification can be avoided, and the effectiveness of stored data is improved.

In one embodiment, a method includes maintaining a first open logical erase block for user writes, maintaining a second open logical erase block for relocate writes, wherein the first and second open logical erase blocks are different logical erase blocks, receiving a first data stream having the user writes, transferring the first data stream to the first open logical erase block, receiving a second data stream having the relocate writes, and transferring the second data stream to the second open logical erase block. Other systems, methods, and computer program products are described in additional embodiments.

Apparatus and methods provide relatively low uncorrectable bit error rates, low write amplification, long life, fast and efficient retrieval, and efficient storage density such that a solid-state drive (SSD) can be implemented using relatively inexpensive MLC Flash for an enterprise storage application. Both primary parity symbols for primary codewords and secondary parity symbols for secondary codewords are generated. The secondary parity symbols are spread out across each page of a group of pages.

Apparatus and methods provide relatively low uncorrectable bit error rates, low write amplification, long life, fast and efficient retrieval, and efficient storage density such that a solid-state drive (SSD) can be reliably implemented using various types of memory cells, including relatively inexpensive multi-level cell flash. One embodiment intelligently coordinates remapping of bad blocks with error correction code control, which eliminates the tables used to avoid bad blocks.

The invention discloses a method for selecting a garbage collection target block, and a solid state disk. The method is characterized in that the recovery factor score of each block is calculated according to the proportion and the erasing frequencies of effective pages in each block, and the block with a lowest score is selected as a recovery target block. The effective page ratio and the erasingfrequency weighting score value of each block on the SSD (Solid State Disk) are calculated to comprehensively consider an influence on the read-write performance and the service life of the SSD as far as possible, and the block with a low effective page proportion and few erasing frequencies is selected as the target recovery block of a GC. A wear-leveling thought is used for selecting an optimalblock as a recovery block, block erasing frequencies in the SSD are reduced, data migration frequencies are lowered, write amplification is reduced, and the service life of the SSD is improved.