Sandisk Technologies Inc.

Western Digital Technologies, Inc.

Vimal Kumar Jain

Ramanathan Muthiah

Dinesh Kumar Agarwal

Vijay Sivasankaran

Aditya Pratap Sharma

Yehuda Hahn

Abhijeet Manohar

Aaron Lee

Eldhose Peter

Rakesh Balakrishnan

Sourabh Sankule

Akhilesh Yadav

Vinod Sasidharan

Niraj Srimal

Adarsh Sreedhar

Anne Pao-Ling Koh

Chetan Agrawal

Bala Siva Kumar Narala

Nagaraj Dandigenahalli Rudrappa

Pradeep Seetaram Hegde

Balasiva Kumar Narala

Kumar Amarjit

Kaushik Kumar Bar

N Balasiva Kumar

A multi-protocol storage device avoids entering a thermal shutdown mode by switching between protocols. The storage device communicates with a host in a first mode using a first protocol. The storage device receives a temperature request from the host, monitors its temperature, and transmits a response to the host when the temperature of the storage device meets a predefined temperature that is below a thermal shutdown threshold. The storage device receives a thermal throttling instruction from the host and switches to a second mode to communicate with the host using a second protocol that uses less resources than the first protocol. The storage device performs thermal throttling until the temperature of the storage device reaches a normal temperature zone. When the temperature of the storage device returns to the normal temperature zone, the storage device returns to the first mode.

Thermal recovery in multi-protocol devices

Some data storage devices have a plurality of memory dies that can be read in parallel for certain types of read requests. Read requests pertaining to a garbage collection operation are often generated sequentially and, thus, are not eligible for parallel execution in the memory dies. In an example data storage device presented herein, such read requests are consolidated and sent to the memory for execution in parallel across the memory dies.

Data storage device and method for improving asynchronous independent plane read (AIPR) utilization

Apparatuses, systems, and methods for low latency parity for a memory device include a controller configured to accumulate, in a memory buffer, combined parity data for a plurality of regions of memory of a memory device in response to write operations for the plurality of regions of memory. The controller is configured to perform a recovery operation for a region of memory in response to determining that a latency setting for the region satisfies a latency threshold. The controller is configured to service a read request for data from the region based on a recovery operation to satisfy the latency setting.

Low latency parity for a memory device

Methods and apparatus are provided for arbitrating access to, and usage of, various device resources of a data storage device (DSD) configured for Machine Learning with Low-Power. The data storage device may include a TinyML controller with an artificial intelligence (AI) accelerator integrated with a data storage controller on a system-on-a-chip (SoC). The device resources may be, e.g., storage resources such as random access memory (RAM) devices, non-volatile memory (NVM) arrays, and latches formed on NVM dies of the NVM arrays. The resource arbitration may be based, for example, on parameters pertaining to ML operations performed by an ML controller that includes the AI accelerator, such as a turnaround time of an ML epoch or a stage-wise execution time. The resource arbitration is configured to provide for the efficient interleaving of the ML/AI operations performed by the ML controller and data storage operations performed by the data storage controller.

Resource usage arbitration in non-volatile memory (NVM) data storage devices with artificial intelligence accelerators

Technology is disclosed for relocating data in a non-volatile storage system. An integrated memory assembly has a control die and a memory die that contains the memory cells. The control die contains control circuitry that relocates data from one set of physical addresses on the memory die to another set of physical addresses on the memory die. This relocation results in a change of a mapping between logical addresses for the data and the physical addresses for the data. The control circuitry may update an L2P table on the memory die after the relocation to map the logical addresses of the data to the second set of physical addresses. The control die may construct a validity bitmap, which specifies whether data at a physical address is valid or invalid. The foregoing reduces data transfer between the integrated memory assembly and a memory controller, which saves time and power.

Data auto-relocation in an integrated memory assembly

A non-volatile storage system includes a memory controller connected to an integrated memory assembly. The integrated memory assembly includes a memory die comprising non-volatile memory cells and a control die bonded to the memory die. The memory controller receives commands from a host, performs logical address to physical address translation ('address translation') operations for the commands, and instructs the integrated memory assembly to perform one or more operations in support of the command. The control die also includes the ability to perform the address translation. When performing a command from the host, the memory controller can choose to perform the necessary address translation or instruct the control die to perform the address translation. When the control die performs the address translation, the resulting physical address is used by the control die to perform one or more operations in support of the command.

Integrated non-volatile memory assembly with address translation

A method, an apparatus, and a system for data address management in non-volatile memory. Write data is allocated to each of a plurality of multi-level pages configured for storage on a page of a non-volatile memory array. A digest is associated with the write data of one multi-level page based on an attribute for that multi-level page. This attribute differs from the attributes of at least one of the other multi-level pages. An amount of redundancy data to be stored with write data on the multi-level page is reduced to account for the associated digest. A digest may be distributed among a plurality of ECC codewords of a multi-level page. The reduced redundancy data, the digest, and the write data for the multi-level page are stored on the page along with the write data for each of the other multi-level pages of the plurality of multi-level pages.

Data address management in non-volatile memory

A system and method for managing multiple file systems on a single non-volatile memory system is described. The system may include a non-volatile memory system with non-volatile memory having first and second file systems, each associated with respective files, and having a common pool of free space. The controller may be configured to update a file system to be mounted to reflect a capacity relating to only the respective files for that file system and all of the common pool of free space, while hiding from the host the file system not being mounted. The method may include the controller only presenting a single file system and hiding the unmounted file system, or may include the controller managing multiple file systems by presenting multiple file systems concurrently.

System and method for managing multiple file systems in a memory

Storage devices, and methods for use therewith, are described herein. Such storage devices can include flash memory, random access memory (RAM), and a memory controller in communication therewith. To improve write performance, the memory controller is configured to store first and second data, corresponding to consecutive unaligned first and second write commands received within a threshold amount of time of one another from a host, sequentially relative to one another within the flash memory. This can involve temporarily storing a tail portion of the first data in the RAM until after a front portion of the first data is stored in the flash memory, and thereafter (after the second write command is received) using the tail portion of the first data to pre-pad a front portion of the second data when the second data is being stored in the flash memory.

Handling of unaligned sequential writes

A storage module may be configured to organize data to be moved from an initial storage location to a destination storage location into sets, and to determine whether to commit the data to the destination storage location on a set-by-set basis. Error correction and/or a post write and read process may be performed on the sets that are copied to the destination storage location to determine whether to commit each of the copied sets.

Moving and committing valid data on a set-by-set basis

A solution for combining a portion of data from a block of single level cells to a block of multi-level cells. The solution includes identifying word lines with only valid data and word lines with non-valid data in a selected block of single level cells, copying data from word lines with valid data to a destination block of multi-level cells and copying data from word lines in the selected block of single level cells with non-valid data to a separate compaction block of single level cells. The system includes a first controller module configured to scan for word lines with only valid data and pass a bitmap identifying valid and invalid word lines to a second controller module. The second controller module is configured to perform on-chip combining of data from valid word lines, and copy data from invalid data word lines to a compaction block of single level cells.

System and method for folding partial blocks into multi-level cell memory blocks

In one embodiment, a memory system is provided comprising a memory die and a controller. The memory die comprises a non-volatile memory, a data latch, and an on-chip randomizer. The controller is configured to send a command to the memory die to cause the on-chip randomizer to store random data in the data latch and send data to the memory die to overwrite some, but not all, of the random data in the data latch, wherein the memory die is configured to transfer the data and random data stored in the data latch to the non-volatile memory. Other embodiments are provided.

Memory system and method for efficient padding of memory pages

A memory system logs failures to optimize garbage collection in partial bad blocks that are reused in non-volatile memory. A failure in a primary block may be logged in an inverse global address table. A garbage collection operation can reference the log in order to automatically avoid the failure in the primary block when the primary block is picked as the source block for garbage collection. Likewise, the garbage collection operation may scan only the logged wordlines in the secondary block when the secondary block is picked as the source block for garbage collection.

Failure logging mechanism to reduce garbage collection time in partially reused bad blocks

A non-volatile storage system comprises a controller and one or more memory die in communication with the controller. The controller sends data and an initial address in conjunction with a request to program the data to one of the memory die. The memory die comprises a plurality of non-volatile storage elements and one or more control circuits. The one or more control circuits attempt to program the data to the non-volatile storage elements at the initial address and determine that programming of the data at the initial address fails. The one or more managing circuits automatically identify a new address in the memory die without the memory die being instructed of the new address by the controller and program the data at the new address on the memory die without the data being re-transmitted from the controller to the memory die.

Growing community of inventors

Bengaluru, India

Vimal Kumar Jain