Pure Storage, Inc.

Google Inc.

Yuhong Mao

Hari Kannan

Robert H Lee

Richard Lee Sites

Peter E Kirkpatrick

Gordon James Coleman

Boris Feigin

Russell Sears

Jeffrey Clifford Mogul

Yijie Zhao

John Colgrove

Ronald S Karr

Ethan L Miller

Uday Ramakrishna Naik

Shuyi Shao

Manoj Kasichainula

John D Davis

Andrew R Bernat

Benjamin Charles Serebrin

Grigori Inozemtsev

Joern W Engel

Jakov Seizovic

Mark Heuchert

John Davis

A method of operating a storage system is provided. The method includes executing an operating system on one or more processors of a compute device that is coupled to one or more solid-state drives and executing a file system on the one or more processors of the compute device. The method includes configuring the compute device with one or more replaceable plug-ins that are specific to the one or more solid-state drives, and executing a flash translation layer on the one or more processors of the compute device, with assistance through the one or more replaceable plug-ins for reading and writing the one or more solid-state drives.

Updates for flash translation layer

A method of storing data is provided. The method includes receiving a plurality of data blocks provided to a hyperscaler system. The method also includes determining a corresponding property for each data block of the plurality of data blocks. The method further includes identifying a set of data blocks from the plurality of data blocks. Each data block of the set of data blocks is associated with a first property. The method further includes storing the set of data blocks in a first zone of a zoned storage system, based on the first property.

Efficient data storage by grouping similar data within a zone

Reading data stored at a free block of a storage device is read prior to allocating the free block for storage of data. A determination as to whether a number of bit flips of the data stored at the free block is below a threshold is made. The free block is added to a pool of active free blocks to be allocated for the storage of data upon determining that the number of bit flips of the data stored at the free block is below the threshold.

Data block allocation for storage system

Intra-block error correction including generating reconstructed sub-portions of a portion of a dataset that is distributed into a plurality of portions that are stored on different storage devices of the storage system, wherein the dataset indicates a first number of errors that is greater than a second number of parity data portions of the dataset, determining that a sub-portion of the dataset does not match a corresponding reconstructed sub-portion of the dataset; and based on the determination, replacing the sub-portion of the dataset with the reconstructed sub-portion of the dataset.

Using intra-block redundancy calculations

A method for processing blocks of flash memory to decrease raw bit errors from the flash memory is provided. The method includes identifying one or more blocks of the flash memory for a refresh operation and writing information regarding the identified blocks, to a data structure. The method includes issuing background reads to the identified blocks, according to the data structure, as the refresh operation. The method may be embodied on a computer readable medium. In some embodiments the background reads may be based on a time based refresh responsive to an increase in raw bit error count in the flash memory over time.

Background reads for solid state storage

A system and method for protocol independent receive side scaling (RSS) includes storing a plurality of RSS hash M-tuple definitions, each definition corresponding to one of a set of possible protocol header combinations for routing an incoming packet, the set of possible protocol header combinations being modifiable to include later-developed protocols. Based on initial bytes of the incoming packet, a pattern of protocol headers is detected, and used to select one of the plurality of RSS hash M-tuple definitions. The selected RSS hash M-tuple definition is applied as a protocol-independent arbitrary set of bits to the headers of the incoming packet to form a RSS hash M-tuple vector, which is used to compute a RSS hash. Based on the RSS hash, a particular queue is selected from a set of destination queues identified for the packet, and the packet is delivered to the selected particular queue.

Protocol-independent receive-side scaling

One or more performance parameters associated with data stored at a storage device of a plurality of storage devices are received by a storage controller. A first number of blocks of the storage device to a high resiliency portion and a second number of blocks of the storage device to a low resiliency portion of the storage device are allocated based on the one or more performance parameters.

Resiliency management in a storage system

A storage cluster with disaggregated compute resources and storage memory is provided. The storage cluster includes a plurality of blades coupled as the storage cluster, each of at least a subset of the plurality of blades having solid-state storage memory therein. The storage cluster includes a switch that direct network-connects a plurality of processors, as compute resources in the plurality of blades, and the solid-state storage memory in each of the at least a subset of the plurality of blades, wherein the compute resources and the solid-state storage memory are disaggregated in the storage cluster.

Disaggregated compute resources and storage resources in a storage system

Refresh of differing capacity NAND

A method of operating a storage unit having non-volatile random-access memory (NVRAM) and solid-state memory is provided. The method includes transferring contents of the NVRAM to the solid-state memory, in response to detecting a power loss. The method includes during the transferring, having each of a plurality of channels in parallel, reading one or more of a plurality of logical unit numbers (LUNs) each corresponding to a portion of the NVRAM, performing an XOR of data of each of the one or more of the plurality of LUNs with data of a preceding LUN, and writing results of the XOR to the solid-state memory.

Efficient memory dump

Intra-block error correction

A storage system, blades, removable modules, and method of configuring a storage system are described. The storage system has blades with computing resources and storage resources. At least one of the blades has, or has added, one or more removable modules.

System having modular accelerators

Optimizing block allocation in a data storage system

A redundant array of independent drives (RAID) stripe is formed across a set of storage controllers of a plurality of storage controllers, wherein the RAID stripe comprises two or more of a plurality of modular storage devices of at least one of the set of storage controllers. The RAID stripe is written across the set of storage controllers.

Optimized data resiliency in a modular storage system

A storage system that supports independent scaling of compute resources and storage resources, the storage system including: one or more chassis, wherein each chassis includes a plurality of slots, each slot configured to receive a blade; a plurality of compute resources; a plurality of storage resources; a plurality of blades, where each blade includes at least one compute resource or at least one storage resource and each of the storage resources may be directly accessed by each of the compute resources without utilizing an intermediate compute resource; a first power domain configured to deliver power to one or more of the compute resources; and a second power domain configured to deliver power to the storage resources, wherein the first power domain and the second power domain can be independently operated.

Growing community of inventors

Fremont, CA, United States of America

Yuhong Mao