Marvell International Limited

Marvell Asia Pte., Ltd.

Marvellworld Trade Ltd.

Nedeljko Varnica

Gregory Burd

Zining Wu

Panu Chaichanavong

Shashi Kiran Chilappagari

Nitin Nangare

Xueshi Yang

Yifei Zhang

Seo-How Low

Kiran Kumar Gunnam

Heng Tang

Dung Viet Nguyen

Mats Oberg

Christophe Therene

Shaohua Yang

Engling Yeo

Phong Sy Nguyen

Noam Mizrahi

Farshid Rafiee Rad

Konstantin Kudryavtsev

Manish Shrivastava

Leo Jiang

Hongwei Song

Lingyan Sun

Viet-Dzung Nguyen

Nirmal V Shende

Xuanxuan Lu

Scott Furey

Ruwan Ratnayake

Igor Polivanyi

Wen Lung Chang

Di Fan

Chin-Jen Pang

Quynh Chau

Sashi Kiran Chilappagari

A solid state drive (SSD) device includes a memory having a plurality of memory cells and an encoder configured to encode information using a polar code to generate encoded information to be stored in the memory. The polar code is constructed based on a plurality of channel models corresponding to different read channel scenarios, including at least a first channel model corresponding a first read channel scenario and a second channel model corresponding to a second read channel scenario, the second read channel scenario different from the first read channel scenario. The SSD device also includes a controller configured to write the encoded information to memory cells in the memory, and read the encoded information from the memory cells in the memory using a selected one of the first read channel scenario and the second read channel scenario.

Polar codes for error correction in non-volatile memory devices

A memory device includes a plurality of memory cells. Each memory cell stores a plurality of signal levels representing a plurality of values corresponding to a respective plurality of bits, bits in corresponding respective positions of significance across the plurality of memory cells constituting respective memory pages of the memory device. The memory device also includes decoding circuitry to decode each bit value of one of the respective memory pages using bit values read from at least one other one of the respective memory pages, adjacent to the one of the respective memory pages. The plurality of signal levels may represent the plurality of values according to a Gray code. The decoding circuitry may be configured to compare each signal level to a set of voltage thresholds, and to decode a subset of the plurality of signal levels using fewer than all voltage thresholds in the set of voltage thresholds.

Adjacent-page-assisted memory decoder and method

A method of reading data read from a NAND Flash memory device includes decoding a set of data read from the device, using an initial set of hard bit thresholds, when the decoding is unsuccessful, performing a read-retry operation that retries the decoding using, in order, each of a plurality of entries in a read-retry table of hard bit thresholds, stopping when decoding based on one of the entries is successful, and when the read-retry operation is unsuccessful, performing a deep retry operation using a set of log-likelihood ratios (LLRs) that vary in at least one of values or symmetries. NAND Flash memory apparatus includes a Flash media controller, a data bus, and an adaptive LLR engine configured to generate, for use in a deep retry operation, a set of LLRs that, and to transfer the set of LLRs that vary to the media controller via the bus.

Adaptive read recovery for NAND flash memory devices

The present disclosure describes apparatuses and methods for implementing a reinforcement learning-enabled low-density parity check (LDPC) decoder. In aspects, an RL-enabled LDPC decoder processes, as part of a first decoding iteration, data of a channel to generate LDPC state information and provides the LDPC state information to a machine learning (ML) algorithm of an RL agent. The RL-enabled LDPC decoder is then configured with LDPC decoding parameters obtained from the ML algorithm and processes, as part of a second decoding operation, the data using the decoding parameters to generate subsequent LDPC state information. The RL-enabled LDPC decoder provides decoded data of the channel based on the subsequent LDPC state information. By using the LDPC decoding parameters provided by the ML algorithm of the RL agent, the RL-enabled LDPC decoder may decode channel data in fewer decoding iterations or with a higher success rate, thereby improving LDPC decoding performance.

Reinforcement learning-enabled low-density parity check decoder

This disclosure describes a storage aggregator controller with metadata computation control. The storage aggregator controller communicates, via a host interface, over a computer network with one or more remote hosts, and also communicates, via a storage device interface, with a plurality of local storage devices, which are separate from the remote host(s) and which have respective non-volatile memories. The storage aggregator controller manages the local storage devices for storage or retrieval of media objects. The storage aggregator controller also governs a selective computation, at aggregator control circuitry or at a storage device controller of one or more of the storage devices, of metadata that defines content characteristics of the media objects that are retrieved from the plurality of storage devices or that are received from the one or more hosts over the computer network for storage in the plurality of storage devices.

Storage aggregator controller with metadata computation control

Embodiments described herein provide improved methods and systems for generating metadata for media objects at a computational engine (such as an artificial intelligence engine) within the storage edge controller, and for storing and using such metadata, in data processing systems.

Storage edge controller with a metadata computational engine

The present disclosure describes apparatuses and methods for artificial intelligence-enabled management of storage media. In some aspects, a media access manager of a storage media system receives, from a host system, host input/output commands (I/Os) for access to storage media of the storage media system. The media access manager provides information describing the host I/Os to an artificial intelligence engine and receives, from the artificial intelligence engine, a prediction of host system behavior with respect to subsequent access of the storage media. The media access manager then schedules, based on the prediction of host system behavior, the host I/Os for access to the storage media of the storage system. By so doing, the host I/Os may be scheduled to optimize host system access of the storage media, such as to avoid conflict with internal I/Os of the storage system or preempt various thresholds based on upcoming idle time.

Artificial intelligence-enabled management of storage media access

Systems and methods for selecting an optimal error recovery procedure for correcting a read error in a solid-state drive are provided. A machine learning model is trained to forecast which error recovery procedure of a plurality of error recovery procedures is most likely to achieve a predetermined goal given a state of a solid-state drive. The predetermined goal is based on at least one of a read latency and a failure rate of the solid-state drive. A current state of the solid-state drive is determined. An error recovery procedure is selected from among the plurality of error recovery procedures by inputting the current state of the solid-state drive into the trained machine learning model, thereby triggering the trained machine learning model to output the selected error recovery procedure. The selected error recovery procedure is executed to recover data from the solid-state drive.

Solid-state drive error recovery based on machine learning

A method for decoding a codeword includes partitioning the codeword into a plurality of component codewords where each of the component codewords comprising a respective plurality of bits. The respective plurality of bits in each of the plurality of component codewords are interleaved. Each of the plurality of interleaved component codewords are decoded along two dimensions to produce (i) a set of first decoding results and (ii) a set of second decoding results. A short error event is then detected and corrected based on (i) the set of first decoding results and (ii) the set of second decoding results.

Methods and systems for short error event correction in storage channel applications

The present disclosure describes apparatuses and methods for artificial intelligence-enabled search of a storage system. In some aspects, a metadata manager of a storage system receives a label of an object that an AI engine detects in data stored in the storage system. The metadata manager creates, in a relational section of a metadata database, an entry for the detected object with an identifier of the label of the detected object and an address of a node corresponding to the detected object. The metadata manager also creates, in a navigational portion of the metadata database and with the address of the detected object, the node that includes a reference to a relative node of another object and a weight of a relationship between the node and the relative node. By so doing, the metadata database may be searched based on weighted relationships between various nodes, thereby enabling contextual or implicit search of data in the storage system.

Artificial intelligence-enabled search for a storage system

A storage control device coupled to a storage device and located remotely from a host device receives media object data from the host device. The storage control device identifies a type of the media object data and select, based on the identified type, a computational model from among a plurality of computational models for use by a computational engine of the storage control device. The computational engine uses the selected computational model to generate metadata describing the media object data. The metadata is stored in the storage device so as to be selectively retrievable from the storage device separately from the media object data.

Systems and methods for generating metadata describing unstructured data objects at the storage edge

Growing community of inventors

San Jose, CA, United States of America

Nedeljko Varnica