Western Digital Technologies, Inc.

Chun Sei Tsai

Carl E Barlow

William Bernard Boyle

Sang Huynh

Thao Hieu Banh

Kenneth J Smith

Anthony L Pei

Chirag C Patel

Dean V Dang

Choo-Bhin Ong

Colin W Morgan

Michael Stephen Rothberg

Thomas Anthony O'Dell

Jenghung Chen

David E Fowler

David E Wachenschwanz

Chandra M Guda

Kent William Gibbons

Hien T Chu

Kenneth J D'Souza

Cheng Fatt Yee

Steven L Webb

Jillian D Passioukov

Don Brunnett

Stella Achtenberg

Gregory B Thelin

Orhan Beker

Donald Joseph Brunnett

Zhi Ning Wang

Jeffrey L Williams

Ofir Pele

Joseph M Viglione

Kexiu Liu

Alexander Babinski

Kameron Kam-Wai Jung

Tom S Chan

Baoliang Zhang

Balvinder Singh

Scott McClure

Ammar A Zayed

Tom Sai-Cheung Chan

Aimamorn Suvichakorn

Heon Ho Chung

Stephany W Kwan

Jonathan Lloyd

Kenneth T Tabacchi

Amit Chattopadhyay

Boon Sien Chew

Johnathan C Lloyd

Kavita S Patil

Peter H Dinh

Dody Suratman

Jahangir John Kasraee

Ngoc Le

Anand Gupta

Krzysztof Gladysz

Kameron Jung

Andrew W Fok

Methods are provided for tactically deploying machine learning operations within existing storage devices without the need for additional capital investment. Machine learning operations are specifically designed to locate and evaluate multiple types of data to complete an operation, including synthesizing missing data. These operations may be processed within a SoC of a storage device as embedded software. Storage devices designed to utilize machine learning methods within existing configurations can include a non-volatile memory for storing data, executable instructions, and a processor to conduct a variety of steps. The steps can include executing a plurality of applications stored in the non-volatile memory, and receiving a request for data, including measurements, from at least one of the applications. The steps can further determine if the requested data is suitable for substitution by an inference and subsequently select at least one machine learning model for generating a suitable inference.

Embedded multi-attribute machine learning for storage devices

A write command is received to write data in a target zone including one or more tracks and it is determined whether a data structure is stored in a memory for counting writes performed in a track of the target zone. If such a data structure is not stored in the memory, at least one Low Repeat Write (LRW) data structure is created for the target zone to count writes in sectors of respective tracks of the target zone since creation of the at least one LRW data structure. In another aspect, a write count for a segment is incremented in a data structure for a track to account for a write in the segment. A current overall Track Interference (TI) write count for the target zone is incremented if the incremented segment write count is the highest write count among respective segment write counts in the data structure.

Dynamic track interference write counting for magnetic disks

A data storage device is disclosed comprising a non-volatile storage medium (NVSM), and control circuitry configured to receive a plurality of access commands, process the plurality of access commands using a customer prediction model to predict a customer of the data storage device, wherein the customer prediction model is trained off-line based on access patterns of a plurality of different customers. The control circuitry then configures access to the NVSM based on the predicted customer.

Data storage device self-configuring based on customer prediction model

A data storage device is disclosed comprising a head actuated over a disk, wherein the head comprises a fly height actuator (FHA) configured to adjust a fly height of the head over the disk. A temperature of the data storage device is measured, and the FHA is controlled as a function of the measured temperature. A first quality metric of the data storage device is measured, and when the first quality metric falls below a threshold, the FHA control as a function of the measured temperature is disabled and the FHA is controlled to decrease the fly height of the head.

Data storage device controlling head fly height based on temperature

A data storage device is disclosed comprising a first head actuated over a first disk surface, and a second head actuated over a second disk surface. During a first revolution of the disk surfaces, a first set of servo sectors are written to the first disk surface and a second set of servo sectors are written to the second disk surface, wherein the first set of servo sectors are circumferentially offset from the second set of servo sectors. During a second revolution of the disk surfaces, at least one of the servo sectors in the first set is read from the first disk surface and at least one of the servo sectors in the second set is read from the second disk surface to control a position of the first head over the first disk surface.

Data storage device bank writing servo sectors for interleaved servo control processing

A data storage device is disclosed comprising a disk comprising a plurality of data tracks each comprising a plurality of data sectors, and a head actuated over the disk. A first frequency preamble is written to a first data sector of a first data track and a second frequency preamble is written to a second data sector of a second data track, wherein the first frequency is different from the second frequency and the first data track is adjacent to the second data track.

Data storage device employing different frequency preambles in adjacent data tracks

A Data Storage Device (DSD) includes a disk with a zone of tracks to be sequentially written. A sequential write pointer is maintained indicating a next physical location for writing data in the zone. When a write command is received to write data in a target region at a physical location behind the sequential write pointer, it is determined whether a write count for each of the target region and an adjacent region before the target region has reached a threshold number of writes. If it is determined that the write counts for both the target region and the adjacent region have not reached the threshold number of writes, the sequential write pointer for the zone is moved to the physical location and the data is written at the physical location to fulfill the write command.

Management of sequentially written data

A device comprises a first storage media and a controller. The controller configured to generate a first operational metric using a first set of operational data at a first frequency and a second operational metric a second set of operational data at a second frequency different than the first frequency; receive a write command to write data to the first storage media; and divert the write command to write data to a second storage media instead of the first storage media when at least one of the first operational metric and the second operational metric indicates that writing data to the first storage media may be harmful to the magnetic rotating disk or result in data loss.

Data storage device with selective write to a first storage media or a second storage media

An apparatus includes: a media; a head configured to write data on the media; a read channel, coupled to the head, configured to detect servo data from the media; and control circuitry, coupled to the read channel, configured to: generate a position error signal (PES), associated with a write operation of an aggressor track, from the servo data, compare a first threshold to the PES for detecting a write unsafe condition, compare a second threshold, within the first threshold, to the PES for detecting a write squeeze condition, and control a refresh operation of a victim track based on at least one of the comparisons.

Electronic system with data refresh mechanism and method of operation thereof

A data storage device is disclosed comprising a head actuated over a disk comprising a plurality of tracks, wherein at least two of the tracks comprise a thermal asperity. Prior to seeking the head from a first track to a second track, for each of at least two of the thermal asperities a minimum seek delay and a maximum seek delay that will cause the head to miss the thermal asperity are computed, and an allowed seek delay range that will cause the head to miss the thermal asperity is computed. When the undelayed seek will cause the head to hit at least one thermal asperity, the seek is delayed by a seek delay determined based on the minimum seek delays, the maximum seek delays, and the allowed seek delay ranges.

Data storage device delaying seek operation to avoid thermal asperities

A disk drive is disclosed comprising a head actuated over a disk comprising a plurality of data tracks. First data is written to a first data track, and the first data is read from the first data track to measure a first quality metric representing a recording quality of the first data. Second data is written to a second data track, and the second data is read from the second data track to measure a second quality metric representing a recording quality of the second data. The first and second quality metrics are processed to generate a first statistical quality metric. A first defect threshold is generated based on the first statistical quality metric, and at least part of the disk is scanned for defects based on the first defect threshold.

Disk drive calibrating defect threshold based on statistical quality metric measured for reference tracks

A data storage device is disclosed comprising a head actuated over a disk comprising a plurality of tracks. A defect scan of at least one of the tracks is executed, and a log entry is generated when a defect is detected. After the defect scan, a verify operation is executed for at least two of the detected defects. A number of times the verify operation detects a false defect is counted, and whether the head is defective is determined based at least partly on the count.

Data storage device screening heads by verifying defects after defect scan

An apparatus includes: a media; a head over the media; a read channel, coupled to the head, configured to extract data from the media; control circuitry, coupled to the read channel, configured to execute a read command; and wherein the read channel is further configured to: generate, based on extracting the data from the media, a data condition indicator, and provide, for use by the control circuitry, the data and the data condition indicator.

Electronic system with data management mechanism and method of operation thereof

A disk drive is disclosed comprising a head actuated over a disk comprising a plurality of tracks, wherein each track comprises a plurality of data sectors. A test pattern is written to at least one data sector in at least one track, and the test pattern is read from the disk to generate a noisy read signal comprising added noise. An estimated data sequence is detected from the noisy read signal, and an amplitude of the added noise is increased until a failure detection metric reaches a metric threshold. A failure threshold is generated based on the amplitude of the added noise when the failure detection metric reaches the metric threshold, and a failure condition of the disk drive is predicted based on the failure threshold.

Disk drive calibrating failure threshold based on noise power effect on failure detection metric

A data storage device can include a magnetic rotating disk, a solid state memory, and a controller. The controller is configured to generate a first operational metric which is updated at a first frequency and a second operational metric which is updated at a second frequency different than the first frequency, receive a write command to write data to the magnetic rotating disk, and divert the write command to write data to the solid state memory instead of the magnetic rotating disk when at least one of the first operational metric and the second operational metric indicates that writing data to the magnetic rotating disk may be harmful to the magnetic rotating disk or result in data loss.

Growing community of inventors

Tustin, CA, United States of America

Chun Sei Tsai