Pure Storage, Inc.

Emc IP Holding Company LLC

Emc Corporation

Matthew D Fleming

Ying Gao

Eric Michael Lemar

Faissal Sleiman

Eric Mueller

Hari Kannan

Nenad Miladinovic

Boris Feigin

Justin Husted

John Roper

Asif Arif Daud

Shao-Ting Chang

Mayank Rajawat

Harsha Mahuli

Janis Stipins

Mohd Fahadullah

Guangyu Shi

Sourav Ghosh

Pavan Rao

Jeffrey Tremaine

Xinyi Shu

Ronald Stuart Steinke

Eric Michael Verwillow

John Boyle

A method of tracking flash memory in a storage system is provided. The method includes initializing a bad blocks threshold value and marking one or more planes or logical unit numbers (LUNs) of flash memory as bad, responsive to determining that bad blocks in the one or more planes or LUNs meet the bad blocks threshold value. The method includes adjusting the bad blocks threshold value, responsive to exceeding a threshold number or rate of retiring planes or LUNs of flash memory, and repeating the marking and the adjusting, with the bad blocks threshold value capped at a maximum threshold value.

Adaptive threshold for bad flash memory blocks

A method of failure detection in a storage system is performed by the storage system. The method includes detecting a failure in a nonvolatile random access memory device that is in or coupled to a storage device having storage memory. The storage system has multiple NVRAM devices and multiple storage devices that have storage memory. The method includes taking a portion or all of the NVRAM device offline. Taking a portion or all of the NVRAM device offline is responsive to detecting the failure. Taking a portion or all of the NVRAM device off-line is while keeping online the storage memory of the storage device, sufficient ones of the NVRAM devices, and sufficient ones of the storage devices to provide reliable access to data and metadata in the storage system.

Intelligent access to a storage device

A method of operating a plurality of blades of a storage system, performed by the storage system, is provided. The method includes writing data stripes across one or more sets of blades of the plurality of blades within resiliency groups, the plurality of blades having computing resources and storage memory, each resiliency group supporting data recovery in case of loss of a specified number of blades of the resiliency group. The method includes transferring data from a first resiliency group to a second resiliency group, responsive to a change in geometry of the storage system.

Resiliency groups

A method of operating a storage system is provided. The method includes detecting differing amounts of storage memory on two or more of a plurality of blades of the storage system and forming a plurality of resiliency groups of blades, based on the differing amounts of storage memory, each resiliency group comprising a subset of the plurality of blades of the storage system supporting data recovery in case of failure of two of the blades of the resiliency group.

Data resiliency with heterogeneous storage

Implementations are provided herein for data deduplication, and more particularly, to post-process data deduplication on a large scale out storage system. Multiple techniques and implementations are disclosed that offer greater efficiency, higher performance, and more stability when performing post-process data deduplication at large scale. Disclosed implementations are based on a process for data deduplication involving four main phases: enumeration, commonality, sharing, and update. Multi-level hashing can be used to identify candidates for deduplication during the enumeration phase, providing a more efficient use of compute resources. In addition, datasets can be phase rotated through the post-process deduplication steps providing a more controllable deduplication environment as well as a more efficient use of resources.

Scalable post-process deduplication

Embodiments are directed towards cloning files in a distributed file system. In response to a request to create file clones, a file cloning application can be employed to generate file clones based on a source file. The file cloning application can generate a new inode for each new file clone and at least one new shadow store. Data blocks from the source file can be transferred to the shadow store. The transferred data blocks can be replaced with reference blocks that point to the data blocks transferred to the shadow store. Likewise, additional reference blocks pointing to data blocks in the shadow store can be generated for the new file clone inodes. Further, the shadow store can maintain reference counters that track the number of reference blocks pointing to each shadow store block. Also, file clones can be generated based on a point-in-time snapshot of the source file.

File clones in a distributed file system

Growing community of inventors

Santa Clara, CA, United States of America

Matthew D Fleming