Abstract: The present disclosure provides a data computing system. The data computing system comprises: a memory, a processor and an accelerator, wherein the memory is communicatively coupled to the processor and configured to store data to be computed and a computed result, the data being written by the processor; the processor is communicatively coupled to the accelerator and configured to control the accelerator; and the accelerator is communicatively coupled to the memory and configured to access the memory according to pre-configured control information, implement a computing process to produce the computed result and write the computed result back to the memory. The present disclosure also provides an accelerator and a method performed by an accelerator of a data computing system. The present disclosure can improve the execution efficiency of the processor and reduce the computing overhead of the processor.

Abstract: To automate time series forecasting machine learning pipeline generation, a data allocation size of time series data may be determined based on one or more characteristics of a time series data set. The time series data may be allocated for use by candidate machine learning pipelines based on the data allocation size. Features for the time series data may be determined and cached by the candidate machine learning pipelines. Predictions of each of the candidate machine learning pipelines using at least the one or more features may be evaluated. A ranked list of machine learning pipelines may be automatically generated from the candidate machine learning pipelines for time series forecasting based upon evaluating predictions of each of the one or more candidate machine learning pipelines.

Abstract: A memory system includes a controller and a flash memory including a plurality of first blocks. The controller writes a value having a first number of bits per memory cell to a plurality of second blocks, and writes a value having a second number of bits per memory cell to a plurality of third blocks among the first blocks. The second number is more than the first number. The controller writes data from a host device to the second blocks and transcribes valid data from the second blocks to the third blocks. The controller controls the number of second blocks in the first blocks according to an order of completion of the data writing to one or more third blocks and an amount of valid data stored in each of the one or more third blocks.

Abstract: Technology described herein can be employed to automatically recommend a tiering policy for data storage of data at a data storage system, such as a cloud storage system. An example method can comprise determining, by a system comprising a processor, context information defining a data storage attribute applicable to data at a cloud storage system. The method can comprise, in response to determining the context information, generating, by the system, a tiering policy defining an element of tiering storage of data at the cloud storage system, wherein the tiering policy is based on the data storage attribute defined by the context information. The method also can comprise, in response to generating the tiering policy, outputting, by the system, the tiering policy to a storage device associated with a customer. The analysis can be performed using an artificial intelligence process, machine learning process or a combination thereof.

Abstract: Systems and methods described herein synchronize events between various components of storage device during the processing of an exception (i.e., an internal error). The storage device can have a plurality of processors which may each coordinate operations on various domains of storage device processing tasks. An exception occurring in one domain may require input and coordination from other domains within the storage device. Each exception may have a list of predetermined steps needed for completion which are coordinated via a series of sync points placed between exception action clusters which perform a series of specific operations until data or processing from another domain is needed to continue processing. The sync points can be utilized to halt processing in one domain until the other domains are in sync and complete one or more exception action operations. In this way, a streamlined and predictable synchronization between domains may occur during an exception.

Abstract: Fragmented data on a storage device may be additionally stored in a cache. A cache weight for determining storage of a data unit can be determined. For example, a computing device can receive storage device characteristics from a storage device. A data unit comprising multiple fragments may be stored on the storage device. The computing device can receive data unit characteristics from the storage device. The computing device can determine a cache weight for the data unit. The computing device may output the cache weight for determining storage of the data unit in a cache.

Abstract: A memory device, a host device and a memory system are provided. The memory device may include a plurality of storage units configured to store data, and at least one device controller configured to, receive a read command from at least one host device and to read data stored in the plurality of storage units in response to the read command, the at least one host device including at least one host memory including a plurality of HPB (high performance boosting) entry storage regions, and provide the at least one host device with a response command, the response command indicating an activation or deactivation of the plurality of HPB entry storage regions, the response command including HPB entry type information which indicates a HPB entry type of the HPB entry storage region.

Abstract: Disclosed are a data processing system comprising: a memory system for providing a host with a memory map segment including map pieces; and the host for storing the memory map segment as a host map segment and converting a logical address into a physical address using the host map segment. The memory system stores changed map pieces in a map cache, inserts the changed map pieces in a response to a first command, and provides the host with the response. The host updates the host map segment based on the changed map pieces. When a read command includes a logical address and a physical address, the memory system accesses a memory device using the physical address of the read command according to whether the logical address of the read command is stored in the map cache.

Abstract: A method is described, which includes receiving, by a memory subsystem controller from a host system, a host read memory command that references a set of logical block addresses associated with a set of transfer units of a memory device. The controller converts the set of logical block addresses to a set of physical block addresses for the set of transfer units; generates a set of device read memory commands based on the physical block addresses, wherein each device read memory command references at least one physical block address; and generates a first aggregated device read memory command based on a first device read memory command and a second read memory command in response to determining that the first device read memory command is associated with the second device read memory command. The controller thereafter transmits the first aggregated device read memory command to the memory device.

Abstract: Various processes for efficiently and effectively managing huge pages include a process for optimizing memory deduplication of huge pages, optimizing the promotion of one or more base pages to one or more huge pages and optimizing memory compaction of a memory space associated with a huge page.

Abstract: In a memory sub-system, data can be received to be stored at a 3-dimensional (3D) memory component in response to a write operation. A first location of a first layer of the 3D memory component is determined at which to store a first portion of the data, where the first layer is within a first logical unit. A second location of a second layer of the 3D memory component is determined at which to store a second portion of the data, where the second layer is within a second logical unit that is different than the first logical unit. The first portion of the data is caused to be stored in first memory cells at the first location within the first layer. The second portion of the data is caused to be stored in second memory cells at the second location within the second layer.

Abstract: This disclosure provides techniques hierarchical address virtualization within a memory controller and configurable block device allocation. By performing address translation only at select hierarchical levels, a memory controller can be designed to have predictable I/O latency, with brief or otherwise negligible logical-to-physical address translation time. In one embodiment, address transition may be implemented entirely with logical gates and look-up tables of a memory controller integrated circuit, without requiring processor cycles. The disclosed virtualization scheme also provides for flexibility in customizing the configuration of virtual storage devices, to present nearly any desired configuration to a host or client.

Abstract: This specification describes methods and systems for accelerating attribute data access for graph neural network (GNN) processing. An example method includes: receiving a root node identifier corresponding to a node in a graph for GNN processing; determining one or more candidate node identifiers according to the root node identifier, wherein attribute data corresponding to the one or more candidate node identifiers are sequentially stored in a memory; and sampling one or more graph node identifiers at least from the one or more candidate node identifiers for the GNN processing.

Abstract: A memory structure having 2m read ports allowing for concurrent access to n data entries can be constructed using three memory structures each having 2m-1 read ports. The three memory structures include two structures providing access to half of the n data entries, and a difference structure providing access to difference data between the halves of the n data entries. Each pair of the 2m ports is connected to a respective port of each of the 2m-1-port data structures, such that each port of the part can access data entries of a first half of the n data entries either by accessing the structure storing that half directly, or by accessing both the difference structure and the structure containing the second half to reconstruct the data entries of the first half, thus allowing for a pair of ports to concurrently access any of the stored data entries in parallel.

Abstract: Systems and methods for securely and remotely storing data in a remote, distributed redundant array of independent drives (RAID) is provided. RAID storage is accomplished through a series of mapped drives, non-routable Internet protocol (IP) addresses, and routable IP addresses. In addition, authorization to access a RAID controller, network address translation (NAT) system, and domain name system (DNS) system may all be separated, increasing security and allowing storage to be securely distributed among a variety of dispersed storage locations.

Abstract: A method for accessing compressed computer memory residing in physical computer memory is disclosed. In the method, compressed memory blocks are represented as sectors, wherein all sectors contain a fixed number of compressed memory blocks, have a fixed logical size in the form of the fixed number of compressed memory blocks, and have varying physical sizes in the form of the total size of data stored in the respective compressed memory blocks.

Abstract: An information processing apparatus includes a first memory configured to retain a request group, based on requests received by a reception interface, a second memory configured to retain a request group, based on requests received by the reception interface, and a first transmission interface configured to transmit the request group retained in the first memory to the first storage and a second transmission interface configured to transmit the request group retained in the second memory to the second storage. The request group transmitted by the first transmission interface and the request group transmitted by the second transmission interface are different request groups.

Abstract: A memory system may include a volatile memory that is configured to store an address data structure that includes a plurality of logical-to-physical address entries. The address data structure may be stored across a plurality of bank groups of the volatile memory. A controller may be configured to store consecutive logical-to-physical address entries across different bank groups. In turn, during read and write operations for data sets associated with consecutive logical addresses, read requests for physical addresses where the data sets are stored may be sent to multiple bank groups and processed by the multiple banks in parallel.

Abstract: An information handling system may include at least one processor; a network interface; and a physical storage resource including a flash translation layer (FTL) operable to provide a mapping between logical storage addresses and physical storage addresses. The information handling system may be configured to: receive a request for a snapshot; for used portions of the physical storage resource, change a metadata identifier from a used status to a snapshot status; prevent deletion of those portions associated with the snapshot status; and transmit, via the network interface, information associated with the portions that are associated with the snapshot status.

Abstract: There is provided a technique to enable a specific IC chip to reliably activate the other IC chips. An information processing apparatus has first to third IC chips and, after loading an activation program for the third IC chip from a first storage unit connected to the first IC chip into a third storage unit connected to the third IC chip, accesses a register and activates the third control unit.