Artificial intelligence (AI) anomaly monitoring in a storage system. The AI anomaly monitoring may include writing commands into a log jointly with the execution of the commands on storage media of a drive. The log includes information regarding the operation of the drive including, at least, the commands. In turn, each drive in the storage system may include an AI processor core that may access the log and apply an AI analysis to the log to monitor for an anomaly regarding the operation of the drive. As each drive in the storage system may use the AI process core to detect anomalies locally to the drive, the computational and network resources needed to employ the AI monitoring may be reduced.

Several embodiments of memory devices and systems having a variable logical memory capacity are disclosed herein. In one embodiment, a memory device can include a plurality of memory regions that collectively define a physical memory capacity and a controller operably coupled to the plurality of memory regions. The controller is configured to advertise a first logical memory capacity to a host device, determine that at least one of the memory regions is at or near end of life, and in response to the determination—send a notification to the host device that a logical memory capacity of the memory device will be reduced and then retire the at least one of the memory regions.

A method for execution by a storage network begins by issuing a decode threshold number of read requests for a set of encoded data slices to a plurality of storage units of a set of storage units and continues by determining whether less than a decode threshold number of read requests has been received in a time window. The method continues by identifying one or more encoded data slices encoded data slices associated with read requests of the decode threshold number of read requests that have not been received and for an encoded data slice of the one or more encoded data slices, issuing a priority read request to a storage unit storing a copy of the encoded data slice. The method then continues by receiving a response from the storage unit storing the copy of the encoded data, where the storage unit storing the copy of the encoded data slice is adapted to delay one or more maintenance tasks in response to the priority read request.

Embodiments are directed to a method and system of forecasting a disk drive survival period in a data storage network, by obtaining operating system data and manufacturer data for the disk drive to create a dataset, screening the dataset to identify a number of features to be selected for model creation, wherein the data set includes censored data and non-censored data, and performing, in an analytics engine, semi-parametric survival analysis on the data set using transfer learning on the model to provide a time-based failure prediction of the disk drive. A graphical user interface provides to a user the failure prediction in one of text form or graphical form.

Dynamic configuration of backups of production volumes based on desired recovery objectives is provided. A system may obtain a recovery point objective (“RPO”) for a particular production volume. The system may initially back up data, written to the production volume, to a storage volume with certain performance parameters. However, if the write operations to the production volume occur at a high enough rate and/or affect a large enough amount of data, there may be a lag in writing that data to the backup volume. The system may monitor the lag with respect to the specified RPO for backup of the production volume. If the lag approaches the RPO, then the system may dynamically change the configuration of the backup volume to better satisfy the RPO.

Storage efficiency driven migration includes: determining a level of similarity between first data stored on a first storage system and second data stored on a second storage system; determining, in dependence upon the level of similarity, that an expected amount of storage space reduction from migrating similar data exceeds a threshold level; and responsive to determining that the expected amount of storage space reduction exceeds the threshold level, initiating a migration of one or more portions of the first data from the first storage system to the second storage system.

A flash memory system may include a flash memory and a circuit for decoding a result of a read operation on the flash memory using a first codeword. The circuit may be configured to generate an estimated codeword based on a result of hard decoding the first codeword and a result of hard decoding a second codeword. The circuit may be further configured to generate soft information based on the hard decoding result of the first codeword and the estimated codeword. The circuit may be further configured to decode the result of the read operation on the flash memory using the soft information.

Embodiments of the invention utilize an optimized key-value storage engine to strike the optimal balance between cloud-cost and performance and supports queries, including updates, lookups, range queries, inserts, and read-modify-writes. Cloud cost is manifested in purchasing both storage and processing resources. The improved approach has the ability to self-design and instantiate holistic configurations given a workload, a cloud budget, and optionally performance goals and a set of Service Level Agreement (SLA) specifications. A configuration reflects an optimized storage engine design in terms of, for example, the individual data structures design (in-memory and on-disk) in the engine as well as their algorithms and interactions, a cloud provider, and the exact virtual machines to be used.

Techniques for estimating raw bit error rate of data stored in a group of memory cells are described. Encoded data is read from a group of memory cells. A first population value is obtained based on a first number of memory cells in the group of memory cells having a read voltage within a first range of read voltages, each read voltage representing one or more bits of the encoded data. An estimated raw bit error rate of the data is determined to satisfy a first threshold. The determination is made using a first trained machine learning model and based in part on the first population value. A first media management operation is initiated in response to the determination that the estimated raw bit error rate satisfies the first threshold.

In some examples, a system aggregates operational metric data of a plurality of storage volumes into aggregated operational metric data groups that correspond to different workload types of workloads for accessing data of a storage system. The system computes an operational metric for a first workload type of the different workload types, the operational metric relating to a resource of the storage system, where the computing of the operational metric for the first workload type comprises inputting aggregated operational metric data of a first aggregated operational metric data group of the aggregated operational metric data groups into a model trained at a system level of the storage system.