Emc IP Holding Company LLC

Dell Products Inc.

Peter Beale

Zachary W Arnold

Bina K Thakkar

Fatemeh Azmandian

Vibhor Kaushik

Owen Martin

Muzhar S Khokhar

Kevin S Labonte

Michael G Varteresian

Charles S Langley

Wenjin Liu

Kshitij Patel

Karthik Hubli

Siva Rama Krishna Kottapalli

Siva Kottapalli

Vijayagomathi Ramasamy

An apparatus comprises a processing device configured to obtain performance data for a plurality of workloads, to select a subset of the performance data corresponding to a subset of the plurality of workloads having a given workload type, and to generate a model characterizing an expected performance of the given workload type by analyzing the selected subset of the performance data to estimate a queuing curve characterizing the expected performance of the given workload type. The processing device is also configured, responsive to determining that a quality of the generated model is above a designated threshold quality level, to utilize the generated model to identify performance impacting events for a given workload of the given workload type and to modify provisioning of compute, storage and network resources allocated to the given workload responsive to identifying performance impacting events for the given workload of the given workload type.

Performance impact detection using queuing curve analysis

Methods, apparatus, and processor-readable storage media for machine learning-based anomaly detection using time series decomposition are provided herein. An example computer-implemented method includes processing, via machine learning techniques pertaining to time series decomposition functions, a first set of historical time series data derived from multiple systems within an enterprise; generating, based on the processed data, one or more pairs of upper bounds and lower bounds directed to system metrics; identifying system anomalies attributed to one or more of the multiple systems within the enterprise by comparing a second set of historical time series data derived from the one or more systems against the one or more pairs of upper bounds and lower bounds; prioritizing, via machine learning techniques pertaining to weighting functions, the system anomalies; and outputting, in accordance with the prioritization, the system anomalies to a user within the enterprise.

Machine learning-based anomaly detection using time series decomposition

Methods, apparatus, and processor-readable storage media for determining similarity between time series using machine learning techniques are provided herein. An example computer-implemented method includes obtaining a primary time series and a set of multiple candidate time series; calculating, using machine learning techniques, similarity measurements between the primary time series and each of the candidate time series; for each of the similarity measurements, assigning weights to the candidate time series based on similarity to the primary time series relative to the other candidate time series; generating, for each of the candidate time series, a similarity score based on the weights assigned to each of the candidate time series across the similarity measurements; and outputting, based on the similarity scores, identification of at least one candidate time series for use in one or more automated actions relating to at least one system.

Determining similarity between time series using machine learning techniques

Recommending configuration changes may include: receiving a decision tree comprising levels of nodes, wherein the decision tree includes leaf nodes each representing a different one of a plurality of hardware configurations, wherein a first leaf represents a first hardware configuration and the first leaf node is associated with a set of I/O workload features denoting a I/O workload of a first system having the first hardware configuration, wherein the set of I/O workload features is associated with an action from the first leaf node to a second leaf node, wherein the second leaf node represents a second hardware configuration and the action represents a hardware configuration change made to transition from the first to the second hardware configuration; and performing processing that determines, using the decision tree, a recommendation for a hardware configuration change for a second system having the first hardware configuration represented by the first leaf node.

Techniques for recommending configuration changes using a decision tree

A method includes obtaining, in a given log processing node, at least two different types of logs associated with assets of an enterprise system. The method also includes generating, at the given log processing node, frequency scores for terms in unstructured log data of each of the different log types, the generated frequency score for a given term in unstructured log data of a given log type being based on (i) occurrence of the given term in historical logs of the given log type previously processed by log processing nodes and (ii) occurrence of the given term in the obtained logs of the given log type. The method further includes extracting, at the given log processing node, features from the obtained logs based on the frequency scores, detecting events affecting the assets utilizing the extracted features, and modifying a configuration of the assets responsive to detecting the events.

Automatic feature extraction from unstructured log data utilizing term frequency scores

Methods, apparatus, and processor-readable storage media for automated stateful counter aggregation of device data are provided herein. An example computer-implemented method includes obtaining historical aggregate counter data and historical individual member counter data associated with a variable set of device members and a given temporal period; computing one or more stateful aggregate counter data values attributed to at least a portion of the variable set of device members for a given temporal value by applying at least one stateful counter aggregation algorithm to the obtained data; and performing one or more automated actions based at least in part on the one or more computed stateful aggregate counter data values.

Automated stateful counter aggregation of device data

Methods, apparatus, and processor-readable storage media for automatic identification of resources in contention in storage systems using machine learning techniques are provided herein. An example computer-implemented method includes obtaining a primary time series and multiple candidate time series; calculating, using machine learning techniques, similarity measurements between the primary time series and each candidate time series; for each similarity measurement, assigning weights to the candidate time series based on similarity to the primary time series relative to the other candidate time series; generating, for each candidate time series, a similarity score based on the weights assigned across the similarity measurements; identifying, based on the similarity scores, at least one of the candidate time series as representative of at least one resource in contention with respect to latency data represented by the primary time series; and outputting identification of the identified candidate time series for use in automated actions.

Automatic identification of resources in contention in storage systems using machine learning techniques

Methods, apparatus, and processor-readable storage media for automatic identification of workloads contributing to behavioral changes in storage systems using machine learning techniques are provided herein. An example computer-implemented method includes obtaining a primary time series and a set of candidate time series; calculating, using machine learning techniques, similarity measurements between the primary time series and each candidate time series in the set; for each similarity measurement, assigning weights to the candidate time series based on similarity values; generating, for each candidate time series, a similarity score based on the assigned weights; automatically identifying, based on the similarity scores, a candidate time series as contributing to an anomaly exhibited in the primary time series; and outputting identifying information of the at least one identified candidate time series for use in one or more automated actions associated with the storage system.

Automatic identification of workloads contributing to behavioral changes in storage systems using machine learning techniques

Methods, apparatus, and processor-readable storage media for automatic identification of workloads contributing to system performance degradation are provided herein. An example computer-implemented method includes obtaining, in connection with a system exhibiting performance degradation, a primary time series and a set of multiple candidate time series; calculating, using machine learning, similarity measurements between the primary time series and each time series in the set; for each measurement, assigning weights to the time series based on similarity to the primary time series relative to the other time series in the set; generating, for each time series in the set, a similarity score based on the weights assigned across the similarity measurements; and outputting, based on the similarity scores, identification of a candidate time series for use in automated actions.

Automatic identification of workloads contributing to system performance degradation using machine learning techniques

A technique manages data storage equipment. The technique involves receiving queue depth metrics from data storage performance data describing data storage performance of the data storage equipment. The technique further involves performing a performance impact detection operation on the queue depth metrics to determine whether a performance impacting event has occurred on the data storage equipment. The technique further involves, in response to a result of the performance impact detection operation indicating that a performance impacting event has occurred on the data storage equipment, launching a set of performance impact operations to address the performance impacting event that occurred on the data storage equipment. Such a technique may be performed by an electronic apparatus coupled with the data storage equipment (e.g., over a network).

Identifying performance impact events in data storage equipment based on queue depth metrics

A method includes obtaining historical storage resource utilization data for a given set of storage resources of one or more storage systems, and generating a plurality of model-specific storage resource capacity predictions utilizing the historical storage resource utilization data and respective ones of a plurality of time series capacity prediction forecasting models. The method also includes selecting a subset of the model-specific storage resource capacity predictions having one or more designated characteristics, determining an overall storage resource capacity prediction based at least in part on a combination of the selected subset of the model-specific storage resource capacity predictions, and modifying a provisioning of storage resources of the one or more storage systems based at least in part on the overall storage resource capacity prediction.

Storage resource capacity prediction utilizing a plurality of time series forecasting models

A method, computer program product, and computing system for receiving telemetry data from a remote storage system. The remote storage system includes a plurality of subsystems. The telemetry data is analyzed to assign a subsystem health score to each of the plurality of subsystems, thus defining a plurality of subsystem health scores. A system health score is assigned to the remote storage system based, at least in part, upon each of the subsystem health scores.

Growing community of inventors

Acton, MA, United States of America

Peter Beale