Abstract: A memory controller can include an error correction module for extended lifetime memory that tracks at least one sized block of non-fault consecutive bits within the disabled page as spare blocks and reuses the spare blocks from the disabled pages as an error correction resource for active blocks. The active blocks can store data, data and metadata, or metadata only (e.g., error correction metadata). A method for extended lifetime memory can include, for an active block of metadata containing at least one fault, using at least one spare block to correct the data of the active block. For an active block of data containing at least one fault, the data can be initially corrected via XOR correction with a first spare block and then ultimately corrected via XOR correction with a second spare block.

Abstract: A printed circuit board (PCB) is provided. The PCB includes a connector extending from a surface of the PCB and a bypass feature extending through the PCB. The PCB is constructed so that a first copy of the PCB is configured to be assembled to a second copy of the PCB with the second copy rotated and/or flipped relative to the first copy. An electrical connection to the first copy is accessible via the connector of the first copy, and an electrical connection to the second copy is accessible via the connector of the second copy through the bypass feature of the first copy.

Abstract: A method for managing processing power in a storage system is provided. The method includes providing a plurality of blades, each of a first subset having a storage node and storage memory, and each of a second, differing subset having a compute-only node. The method includes distributing authorities across the plurality of blades, to a plurality of nodes including at least one compute-only node, wherein each authority has ownership of a range of user data.

Abstract: A codeword is generated from a message. One or more anchor values are appended to the codeword at predetermined anchor positions. Before the codeword is stored in a memory block, the locations and values of stuck cells in the memory block are determined. Based on the values and positions of the stuck cells, the values of the codeword are remapped so that values of the codeword that are the same as the values of the stuck cells are placed at the positions of the stuck cells. The remapped codeword is stored in the memory block. When the message is later read, the original codeword can be recovered from the remapped codeword based on the locations of the anchor values in the remapped codeword.

Abstract: A storage system is provided. The storage system includes a plurality of storage units, each of the plurality of storage units having storage memory for user data and a plurality of storage nodes, each of the plurality of storage nodes configured to have ownership of a portion of the user data. The storage system includes a first pathway, coupling the plurality of storage units such that each of the plurality of storage units can communicate with at least one other of the plurality of storage units via the first pathway without assistance from the plurality of storage nodes.

Abstract: Embodiments include a high frequency write through memory device including a plurality of memory cells and a plurality of local evaluation circuits. Each of the plurality of local evaluation circuits are coupled to at least one of the plurality of memory cells and are configured to prevent data stored in the coupled memory cells from being written to a latch node during a write through operation.

Abstract: A universal single-bitstream FPGA library or ASIC implementation accelerates matrix-vector multiplication processing multiple matrix encodings including dense and multiple sparse formats. A hardware-optimized sparse matrix representation referred to herein as the Compressed Variable-Length Bit Vector (CVBV) format is used to take advantage of the capabilities of FPGAs and reduce storage and bandwidth requirements across the matrices compared to that typically achieved when using the Compressed Sparse Row (CSR) format in typical CPU- and GPU-based approaches. Also disclosed is a class of sparse matrix formats that are better suited for FPGA implementations than existing formats reducing storage and bandwidth requirements. A partitioned CVBV format is described to enable parallel decoding.

Abstract: A plurality of storage nodes in a single chassis is provided. The plurality of storage nodes in the single chassis is configured to communicate together as a storage cluster. Each of the plurality of storage nodes includes nonvolatile solid-state memory for user data storage. The plurality of storage nodes is configured to distribute the user data and metadata associated with the user data throughout the plurality of storage nodes such that the plurality of storage nodes maintain the ability to read the user data, using erasure coding, despite a loss of two of the plurality of storage nodes. A plurality of compute nodes is included in the single chassis, each of the plurality of compute nodes is configured to communicate with the plurality of storage nodes. A method for accessing user data in a plurality of storage nodes having nonvolatile solid-state memory is also provided.

Abstract: A method of implementing a write block read function for a memory device includes configuring a dynamic read address decoder to receive static read address bits as inputs thereto and to generate an output used to implement a read operation of a memory location corresponding to the read address bits; configuring a dynamic write address decoder to receive static write address bits as inputs thereto and to generate an output used to implement a write operation of a memory location corresponding to the write address bits; and configuring a static write address decoder, in parallel with the dynamic write address decoder, to receive a portion of the static write address bits as inputs thereto, and coupling the static write address decoder to the dynamic read address decoder so as to block the read operation upon an address conflict with the write operation.

Abstract: A write block read apparatus for a memory device includes a dynamic read address decoder that receives static read address bits as inputs thereto and having an output used to implement a read operation of a memory location corresponding to the read address bits; a dynamic write address decoder that receives static write address bits as inputs thereto and having an output used to implement a write operation of a memory location corresponding to the write address bits; and a static write address decoder, configured in parallel with the dynamic write address decoder, the static write address decoder configured to receive a portion of the static write address bits as inputs thereto, and wherein the static write address decoder is coupled to the dynamic read address decoder so as to block the read operation upon an address conflict with the write operation.

Abstract: A codeword is generated from a message. One or more anchor values are appended to the codeword at predetermined anchor positions. Before the codeword is stored in a memory block, the locations and values of stuck cells in the memory block are determined. Based on the values and positions of the stuck cells, the values of the codeword are remapped so that values of the codeword that are the same as the values of the stuck cells are placed at the positions of the stuck cells. The remapped codeword is stored in the memory block. When the message is later read, the original codeword can be recovered from the remapped codeword based on the locations of the anchor values in the remapped codeword.

Abstract: A storage cluster is provided. The storage cluster includes a plurality of storage nodes, each of the plurality of storage nodes having nonvolatile solid-state memory and a plurality of operations queues coupled to the solid-state memory. The plurality of storage nodes is configured to distribute the user data and metadata throughout the plurality of storage nodes such that the plurality of storage nodes can access the user data with a failure of two of the plurality of storage nodes. Each of the plurality of storage nodes is configured to determine whether a read of 1 or more bits in the solid-state memory via a first path is within a latency budget. The plurality of storage nodes is configured to perform a read of user data or metadata via a second path, responsive to a determination that the read of the bit via the first path is not within the latency budget.

Abstract: In some embodiments, a method for die-level monitoring is provided. The method includes distributing user data throughout a plurality of storage nodes through erasure coding, wherein the plurality of storage nodes are housed within a chassis that couples the storage nodes. Each of the storage nodes has a non-volatile solid-state storage with non-volatile memory and the user data is accessible via the erasure coding from a remainder of the storage nodes in event of two of the storage nodes being unreachable. The method includes producing diagnostic information that diagnoses the non-volatile memory on a basis of per package, per die, per plane, per block, or per page, the producing performed by each of the plurality of storage nodes. The method includes writing the diagnostic information to a memory in the storage cluster.

Abstract: A method for adjustable error correction in a storage cluster is provided. The method includes determining health of a non-volatile memory of a non-volatile solid-state storage unit of each of a plurality of storage nodes in a storage cluster on a basis of per flash package, per flash die, per flash plane, per flash block, or per flash page. The determining is performed by the storage cluster. The plurality of storage nodes is housed within a chassis that couples the storage nodes as the storage cluster. The method includes adjusting erasure coding across the plurality of storage nodes based on the health of the non-volatile memory and distributing user data throughout the plurality of storage nodes through the erasure coding. The user data is accessible via the erasure coding from a remainder of the plurality of storage nodes if any of the plurality of storage nodes are unreachable.

Abstract: A storage cluster is provided. The storage cluster includes a plurality of storage nodes within a chassis. The plurality of storage nodes has flash memory for storage of user data and is configured to distribute the user data and metadata throughout the plurality of storage nodes such that the storage nodes can access the user data with a failure of two of the plurality of storage nodes. Each of the storage nodes is configured to generate at least one address translation table that maps around defects in the flash memory on one of a per flash package basis, per flash die basis, per flash plane basis, per flash block basis, per flash page basis, or per physical address basis. Each of the plurality of storage nodes is configured to apply the at least one address translation table to write and read accesses of the user data.

Abstract: A plurality of storage nodes within a single chassis is provided. The plurality of storage nodes is configured to communicate together as a storage cluster. The plurality of storage nodes has a non-volatile solid-state storage for user data storage. The plurality of storage nodes is configured to distribute the user data and metadata associated with the user data throughout the plurality of storage nodes, with erasure coding of the user data. The plurality of storage nodes is configured to recover from failure of two of the plurality of storage nodes by applying the erasure coding to the user data from a remainder of the plurality of storage nodes. The plurality of storage nodes is configured to detect an error and engage in an error recovery via one of a processor of one of the plurality of storage nodes, a processor of the non-volatile solid state storage, or the flash memory.

Abstract: A method of failure mapping is provided. The method includes distributing user data throughout a plurality of storage nodes through erasure coding, wherein the plurality of storage nodes are housed within a chassis that couples the storage nodes as a storage cluster. Each of the plurality of storage nodes has a non-volatile solid-state storage with flash memory or other types of non-volatile memory and the user data is accessible via the erasure coding from a remainder of the plurality of storage nodes in event of two of the plurality of storage nodes being unreachable. The method includes determining that a non-volatile memory block in the memory has a defect and generating a mask that indicates the non-volatile memory block and the defect. The method includes reading from the non-volatile memory block with application of the mask, wherein the reading and the application of the mask are performed by the non-volatile solid-state storage.

Abstract: A memory controller can include an error correction module for extended lifetime memory that tracks at least one sized block of non-fault consecutive bits within the disabled page as spare blocks and reuses the spare blocks from the disabled pages as an error correction resource for active blocks. The active blocks can store data, data and metadata, or metadata only (e.g., error correction metadata). A method for extended lifetime memory can include, for an active block of metadata containing at least one fault, using at least one spare block to correct the data of the active block. For an active block of data containing at least one fault, the data can be initially corrected via XOR correction with a first spare block and then ultimately corrected via XOR correction with a second spare block.

Abstract: A storage system is provided. The storage system includes a plurality of storage units, each of the plurality of storage units having storage memory for user data and a plurality of storage nodes, each of the plurality of storage nodes configured to have ownership of a portion of the user data. The storage system includes a first pathway, coupling the plurality of storage units such that each of the plurality of storage units can communicate with at least one other of the plurality of storage units via the first pathway without assistance from the plurality of storage nodes.

Abstract: A plurality of storage nodes in a single chassis is provided. The plurality of storage nodes in the single chassis is configured to communicate together as a storage cluster. Each of the plurality of storage nodes includes nonvolatile solid-state memory for user data storage. The plurality of storage nodes is configured to distribute the user data and metadata associated with the user data throughout the plurality of storage nodes such that the plurality of storage nodes maintain the ability to read the user data, using erasure coding, despite a loss of two of the plurality of storage nodes. A plurality of compute nodes is included in the single chassis, each of the plurality of compute nodes is configured to communicate with the plurality of storage nodes. A method for accessing user data in a plurality of storage nodes having nonvolatile solid-state memory is also provided.