Seagate Technology Incorporated

Other

Daniel John Benjamin

Ryan James Goss

David W Claude

Ryan Charles Weidemann

Graham David Ferris

Mark Ish

Kevin Arthur Gomez

David Scott Ebsen

Thomas V Spencer

Matthew B Lovell

Steven Scott Williams

Stacey Secatch

Stephen H Perlmutter

Robert Wayne Moss

Benjamin J Scott

Siddhartha Kumar Panda

A data storage system may have a plurality of memory cells located in different data storage devices that are arranged into a plurality of logical namespaces with each logical namespace configured to be sequentially written and entirely erased as a single unit. An asymmetry strategy may be proactively created with the asymmetry module in response to data access activity to the logical namespaces by the asymmetry module. A new mode, as prescribed by the asymmetry strategy, is entered for at least one logical namespace in response to an operational trigger being met. The new mode changes a timing of at least one queued data access request to at least one logical namespace.

Data storage system with workload-based asymmetry compensation

A data storage system may have a data storage device with a memory arranged into a plurality of logical namespaces. A power module can be connected to the plurality of logical namespaces and configured to transition at least one memory cell in response to a workload computed for a namespace of the plurality of the logical namespaces to maintain a power consumption of 8 watts or less for the data storage device.

Data storage system with workload-based dynamic power consumption

A system can log data access activity to a memory array with a metadata module while the memory array is logically divided into multiple namespaces. A workload can be determined for each namespace by the metadata module and a metadata strategy can be created with the metadata module in view of the respective namespace workloads. A first metadata and second metadata may be generated for respective first and second user-generated data for storage into a first namespace of the multiple namespaces. The first metadata can be compressed with a compression level prescribed by the metadata strategy in response to a detected or predicted workload to the first namespace before the first metadata, second metadata, first user-generated data, and second user-generated data are each stored in the first namespace.

Lossless namespace metadata management system

Method and apparatus for enhancing performance of a storage device, such as a solid-state drive (SSD). In some embodiments, the storage device monitors a rate at which client I/O access commands are received from a client to transfer data with a non-volatile memory (NVM) of the storage device. A ratio of background access commands to the client I/O access commands is adjusted to maintain completion rates of the client I/O access commands at a predetermined level. The background access commands transfer data internally with the NVM to prepare the storage device to service the client I/O access commands, and can include internal reads and writes to carry out garbage collection and metadata map updates. The ratio may be adjusted by identifying a workload type subjected to the storage device by the client.

Client input/output (I/O) access rate variation compensation

Apparatus and method for managing metadata in a data storage device such as a solid-state drive (SSD). The metadata are stored in combined (combo) pages in a non-volatile memory (NVM) each having first and second level map entries. The second level map entries provide a logical-to-physical address translation layer for user data blocks stored to the NVM, and the first level map entries describe the second level map entries in the combo page. A global map structure is accessed to identify a selected combo page in the NVM associated with a pending access command. The first and second level map entries are retrieved from the combo page, and the second level map entries are used to identify a target location for the transfer of user data blocks to or from the NVM.

Combined page footer for parallel metadata storage

Apparatus and method for object storage, such as a solid-state drive (SSD) or array thereof. In some embodiments, data arranged as an object are presented for storage to a non-volatile memory (NVM) of a data storage device. Prior to storage, a configuration of the NVM is adaptively adjusted, such as by adjusting a garbage collection unit (GCU) layout, an error correction code (ECC) scheme, and/or a map metadata format used by the NVM. The object is thereafter stored to the NVM using the adaptively adjusted configuration. A controller of the data storage device generates a predicted remaining storage capacity of the NVM in terms of additional objects that can be stored by the NVM responsive to the adaptively adjusted configuration of the NVM. A non-linear sliding scale may be used such that a greater number of smaller objects or a smaller number of larger objects may be accommodated.

Data storage device with adaptive object storage capabilities

Systems and methods for limiting performance variation in a storage device are described. Storage devices receive work requests to perform one or more operations from other computing devices, such as a host computing device. Completing the work requests may take a response time. In some embodiments, if the response time of executing the work request exceeds a threshold, the storage device may assign additional computing resources to complete the work request.

Technologies for limiting performance variation in a storage device

A data storage system can transfer user-generated data from a data storage device to a host responsive to a host workload that consists of data read requests from the host received at a requested rate. A completion rate of a data read request can be delayed in accordance to a data access uniformity strategy to nominally match the requested rate.

Data storage system with access completion uniformity

Method and apparatus for enhancing performance of a storage device, such as a solid-state drive (SSD). A non-volatile memory (NVM) stores user data from a client device. Map metadata in a local memory describes locations of the user data in the NVM. The map metadata is arranged as a snapshot and accumulated journal updates. A metadata manager circuit combines a first portion of the journal updates with the existing snapshot to generate a new snapshot, and places a second portion of the journal updates into a read-only journal table in accordance with a selected metadata journaling strategy. A controller uses the updated snapshot and the table to service subsequently received client commands. Only dirty entries are processed and are written at the slowest acceptable rate, thereby improving client I/O performance during normal operation and time to ready (TTR) performance of the device during initialization.

Processing map metadata updates to reduce client I/O variability and device time to ready (TTR)

Method and apparatus for managing data in a storage device, such as a solid-state drive (SSD). In some embodiments, a first read command from a command queue is forwarded to a non-volatile memory (NVM) to request retrieval of a first set of readback data. While the NVM initiates in-process execution of the first read command, an expanded read command is issued to the NVM. The expanded read command supercedes the first read command and requests an expanded set of readback data that includes the first set of readback data as well as a second set of readback data. The second set of readback data may be associated with a second read command in the command queue. The NVM transfers the expanded set of readback data to a read buffer responsive to the expanded read command. The first and second read commands may be client reads, background reads or both.

Combining in-process reads to reduce die collisions

A system may consist of a network controller connected between a non-volatile memory and a host with a disturb module connected to the non-volatile memory. A received data write request from a host may prompt the assignment of a first physical block address in the non-volatile memory to a data block of the data write request with the network controller. The data block and first physical block address can each be characterized with the disturb module before the first physical block address is altered to a second block address in the non-volatile memory in response to a disturb strategy generated by the disturb module. The second block address can be selected based on the characterization of the data block and first physical block address with respect to a risk of a disturb condition occurring in the non-volatile memory.

Data storage system with read disturb control strategy whereby disturb condition can be predicted

Method and apparatus for enhancing power cycle performance of a storage device, such as a solid-state drive (SSD). In some embodiments, map data that describe the contents of a non-volatile memory (NVM) are arranged as snapshots and intervening journal updates. During a scram interval in which the storage device transitions to a powered down condition, the snapshots and journal updates for primary segments with high client interest are updated prior to storage to the NVM. During a reinitialization interval in which the storage device transitions to a powered up condition, the updated primary segments are loaded, after which the storage device provides the client device with an operationally ready notification. Remaining secondary segments are updated and loaded after the notification. The primary segments are identified based on a detected workload from the client device. Configuration changes can further be made based on the detected workload.

Storage device with enhanced time to ready performance

Method and apparatus for managing data in a storage device, such as a solid-state drive (SSD). In some embodiments, write and read commands from a client device are placed into a command queue pending servicing to transfer data between the client device and a non-volatile memory (NVM). A write cache temporarily stores sets of writeback data pending transfer. A cache manager detects an overlap condition in which a subsequently received command at least partially overlaps a pending write command. In response, the cache manager enacts a change in caching policy that includes retention of the cached writeback data to aid in the servicing of the subsequently received command. The changes in caching policy can include an increase in the size of the write cache, delays in the writing of hot writeback data sets, the coalescing of different writeback data sets, cache hits using the cached writeback data, etc.

Managing overlapping reads and writes in a data cache

Apparatus and method for managing metadata in a data storage device such as a solid-state drive (SSD). In some embodiments, a write stream is formed of user data blocks to be sequentially written to a non-volatile memory (NVM). An entry of a reverse directory footer is generated for each user data block in the write stream to describe a physical address in the NVM at which the corresponding user data block is to be stored. The entries are accumulated in a buffer memory until the total count of entries reaches a predetermined threshold and a complete footer data structure is formed. The complete footer data structure is thereafter inserted into the write stream for writing, with the data blocks, to the NVM. The complete footer data structure has an overall size that corresponds to an overall size of each of the user data blocks.

Reverse directory structure in a garbage collection unit (GCU)

Systems and methods are disclosed for enhanced garbage collection operations at a memory device. The enhanced garbage collection may include selecting data and blocks to garbage collect to improve device performance. Data may be copied and reorganized according to a data stream via which the data was received, or data and blocks may be evaluated for garbage collection based on other access efficiency metrics. Data may be selected for collection based on sequentiality of the data, host access patterns, or other factors. Processing of host commands may be throttled based on a determined amount of work to garbage collect a plurality of blocks, in order to limit variability in host command throughput over a time period.

Growing community of inventors

Savage, MN, United States of America

Daniel John Benjamin