International Business Machines Corporation

Dell Products Inc.

Emc IP Holding Company LLC

Axerra Networks, Inc.

Osnat Shasha

Gregory Tad Kishi

Joseph W Dain

Christopher Zaremba

Rivka Mayraz Matosevich

Michael Keller

Zah Barzik

Shay Haim Akirav

Lior Chen

Jonathan Amit

Ofer Leneman

Shaul Dar

Alon Marx

Dan Cohen

Lior Shlomov

Ariel Waizel

Ram Elron

Anna Stopel

Constantine Gavrilov

Yariv Bachar

Yaacov Frank

Amit Margalit

Vladislav Drouker

Yuval A Ben-Horin

Alex Tal

Tomer Goldberg

Ilya Tkachevsky

Ziv Barak

Alex Kul

Alex Kulakovsky

Daniel M Aizer

Shlomit Neufeld

Alon Shtern

Guy Kronenthal

Raz Korn

An apparatus comprises a processing device configured to determine information characterizing one or more errors detected on a first one of a set of two or more storage systems, the determined information comprising a configuration of the first storage system, error messages associated with the detected errors, recovery actions taken on the first storage system in response to the error messages, and system state information for the first storage system before and after the recovery actions. The processing device is also configured to generate, utilizing one or more machine learning algorithms based at least in part on the determined information, at least one self-healing policy specifying at least one recovery action to take in response to at least one of the error messages. The processing device is further configured to provision the generated at least one self-healing policy in storage controllers of each of the two or more storage systems.

Machine learning-based generation of self-healing policies for error messages encountered on storage systems

Techniques are disclosed for machine learning optimized storage system simulation in a virtual environment. For example, a system includes at least one processing device including a processor coupled to a memory; the at least one processing device being configured to implement the following steps: receiving a series of action alerts from a virtual reality system concerning a virtual reality representation of a storage system; translating each action alert in the action alert series into a corresponding storage system simulator event state, to generate a series of event states; using a machine learning model to determine a new event state based on the series of event states; generating a new event based on the new event state and on the series of event states; and updating a storage system simulation corresponding to the virtual reality representation of the storage system to display the new event in the virtual reality system.

Machine learning optimized storage system simulation in a virtual environment

Techniques are provided for managing quality of service (QoS) policies in a virtual storage environment. A data storage system receives a request from a host system to connect to a target virtual volume in a storage array of the data storage system, and determines a protocol endpoint which has an assigned QoS policy which corresponds to a QoS policy of the target virtual volume. The data storage system binds the target virtual volume to the protocol endpoint which is determined to have an assigned QoS policy which corresponds to the QoS policy of the target virtual volume, and sends a unique identifier of the protocol endpoint, which is bound to the target virtual volume, to the host system. The data storage system utilizes the protocol endpoint, which is bound to the target virtual volume, to handle data access requests received from the host system for accessing the target virtual volume.

Enforcing quality of service in a virtualized storage environment

Provided are a computer program product, system, and method for prefetching data blocks from a primary storage to a secondary storage system while data is being synchronized between the primary storage and secondary storage. A determination is made of data blocks to prefetch from the primary storage to the secondary controller not yet synchronized from the primary storage to the secondary storage in anticipation of future access requests for the data blocks to the secondary controller while data blocks are being synchronized between the primary storage and the secondary storage over the network. A prefetch command is sent to prefetch the determined data blocks to copy from the primary storage to the secondary controller to make available to future access requests received at the secondary controller for the determined data blocks.

Prefetching data blocks from a primary storage to a secondary storage system while data is being synchronized between the primary storage and secondary storage

An apparatus for detecting a degraded grid controller includes a status module that collects for individual grid controllers in a grid array a status update comprising a keepalive status and a connectivity indicator and/or a pending transaction indicator between pairs of the individual grid controllers. The apparatus also includes a responsiveness module that determines a responsiveness indicator for a particular grid controller based on the keepalive status and the connectivity indicator and/or the pending transaction indicator of the particular grid controller. The apparatus further includes a configuration module that removes the particular grid controller of the individual grid controllers from use in the grid array in response to the responsiveness module determining that the responsiveness indicator of the particular grid controller indicates that the particular grid controller is degraded. A method and a system for detecting a degraded grid controller perform similar functions to those of the apparatus.

Apparatus, method, system for detecting a degraded grid controller

Aspects of the present disclosure relate to transmitting prioritized path data to a device based on a set of topology rules and metrics associated with a storage target. A storage target path discovery request is received from the device. In response to the storage target path discovery request, metrics associated with the storage target are collected. Based on the collected metrics, available paths to the storage target are identified. A set of topology rules are then determined. Based on the topology rules and the collected metrics, a subset of available paths are selected. The subset of available paths are then prioritized into prioritized path data. The prioritized path data is then stored and transmitted to the device.

Distributed storage path configuration

Methods, computing systems and computer program products implement embodiments of the present invention that include detecting, on a communications network in a computing facility including a first computer coupled to a second computer via the communications network, an input/output (I/O) request conveyed from the first computer to the second computer. Subsequent to detecting the I/O request, data traffic resulting from the second computer processing the I/O request is identified on the communications network, and network statistics are collected based on the I/O request and the data traffic. Upon detecting a performance degradation of the computing facility, the performance degradation greater than a predetermined amount, heuristics can be applied to the collected network statistics, to identify a cause of the performance degradation and to identify a remedy to the detected performance degradation. In some embodiments, the identified cause and the identified remedy can be presented to a user on a display.

Data traffic monitoring tool

In a hash-based multi-tenancy in a deduplication system, incorporating, as if part of input data, a tenant identification (ID) into a hash value calculation using a single hash based index table for separating data segments in a multi-tenant deduplication system.

Hash-based multi-tenancy in a deduplication system

Provided are a computer program product, computer system, and method for smart selection of a storage module to be excluded when a connection between two storage modules is broken. An indication is received from a first storage module that a connection between the first storage module and a second storage module is broken. In response to determining that the second storage module is accessible, values of exclusion criteria for the first storage module are determined and summed to identify a first exclusion total. Then, values of exclusion criteria for the second storage module are determined and summed to identify a second exclusion total. In response to determining that the first exclusion total exceeds the second exclusion total, the second storage node is excluded from the cluster. In response to determining that the second exclusion total exceeds the first exclusion total, the first storage node is excluded from the cluster.

Smart selection of a storage module to be excluded

Methods, computing systems and computer program products implement embodiments of the present invention that include detecting a loss of communication with a given storage node among multiple storage nodes in a distributed computing system. Upon detecting the loss of communication, a log including updates to the data stored in the given storage node is recorded and, the recorded updates can be applied to the given storage node upon communication with the given storage node being reestablished. In some embodiments, the distributed storage system may be configured as a software defined storage system where the storage nodes can be implemented as either virtual machines or software containers. In additional embodiments, upon detecting the loss of communication, a redistribution of the mirrored data among remaining storage nodes is initiated upon detecting the loss of communication, and the redistribution is rolled back upon reestablishing the communication.

Rebuild rollback support in distributed SDS systems

Methods, computing systems and computer program products implement embodiments of the present invention that include configuring distributed storage system resources for a distributed storage system. Examples of the storage system resources include a plurality of network segments, one or more network devices coupled to the network, and multiple nodes coupled to the network, the nodes including both frontend and backend nodes. Upon receiving, by a given frontend node in the distributed storage system, an input/output (I/O) request, one or more of the distributed storage system resources required to process the I/O request are identified, and a respective load that the I/O request will generate on each of the identified distributed storage system resources is calculated. The distributed storage system processes the I/O request upon detecting that the respective loads are less than respective available capacities of the identified distributed storage system resources.

Preventing input/output (I/O) traffic overloading of an interconnect channel in a distributed data storage system

Methods, computing systems and computer program products implement embodiments of the present invention that include mirroring, in a distributed storage system having multiple storage nodes, data on the storage nodes. Upon the distributed storage system detecting a loss of communication with a given storage node, a log including updates to the data stored in the given storage node is recorded and, the recorded updates can be applied to the given storage node upon communication with the given storage node being reestablished. In some embodiments, the distributed storage system may be configured as a software defined storage system where the storage nodes can be implemented as either virtual machines or software containers. In additional embodiments, upon detecting the loss of communication, a redistribution of the mirrored data among remaining storage nodes is initiated upon detecting the loss of communication, and the redistribution is rolled back upon reestablishing the communication.

Growing community of inventors

Holon, Israel

Osnat Shasha