Dell Products Inc.

Emc IP Holding Company LLC

Shaul Dar

Ramakanth Kanagovi

Guhesh Swaminathan

Rajan Kumar

Vamsi K Vankamamidi

Avitan Gefen

Vladimir Shveidel

Owen Martin

Amitai Alkalay

Amihai Savir

Osnat Shasha

Paras Pandya

Swati Smita Sitha

Sanjib Mallick

Uri Shabi

Ronen Gazit

Gajanan S Natu

Vishnu Murty Karrotu

Michael Zeldich

Haiyun Bao

Anil Kumar Koluguri

David Sydow

Denis Dovzhenko

Alex Kul

Alex Kulakovsky

Ranjith Reddy Basireddy

Rajesh Alevoor Kini

A method, computer program product, and computing system for generating a plurality of chunks for a plurality of text portions of a document, wherein the document includes the plurality of text portions and a plurality of images. Each chunk is indexed using a word embedding. Each of the plurality of images is indexed based upon, at least in part, a position of a respective image relative to a corresponding chunk. An image placeholder is generated for each of the plurality of images. A plurality of image-enhanced embeddings is generated by inserting the image placeholder for each of the plurality of images into a respective word embedding for the corresponding chunk. The plurality of image-enhanced embeddings are provided for processing a query using a generative artificial intelligence (AI) model.

System and method for enhancing generative artificial intelligence (AI) model-based document search with image retrieval

A method, computer program product, and computing system for identifying a plurality of headings from a document by processing a hierarchical structure associated with the document including the plurality of headings and a plurality of content portions within the plurality of headings. A plurality of respective chunks are generated using the plurality of headings and a prompt size limitation associated with a prompt of a generative artificial intelligence (AI) model. The plurality of respective chunks are provided for generating a prompt for the generative AI model.

System and method for generating optimized chunks for retrieval augmented generation using document hierarchy

A method, computer program product, and computing system for processing a plurality of input/output (IO) requests associated with a storage object in a storage system. A plurality of IO features are generated using the plurality of IO requests associated with the storage object. A time dependent IO feature is identified from the plurality of IO features. A coefficient for the time dependent IO feature for the storage system is extracted. The time dependent IO feature is calibrated using the coefficient for the time dependent IO feature for the storage system relative to the time dependent IO feature from at least one other storage system.

System and method for calibrating heterogenous workloads in a machine learning model for storage objects

A method, computer program product, and computing system for processing a plurality of historical input/output (IO) requests associated with a plurality of storage objects of a storage system from a plurality of time intervals. The plurality of storage objects may be divided into a plurality of storage activity classes using a classification-based machine learning model and the plurality of historical IO requests. A next access time for each storage object may be forecasted based upon, at least in part, the plurality of storage activity classes.

System and method for machine learning-based data archiving in storage systems

A method, computer program product, and computing system for monitoring for a potential ransomware attack on a storage object of a storage system based upon, at least in part, processing of a plurality of input/output (IO) features associated with the storage object using a machine learning model. A host computing device associated with the storage object under the potential ransomware attack is identified. A process executing on the host computing device associated with the storage object under the potential ransomware attack is identified. A remedial action is performed on the storage system in response to identifying the process executing on the host computing device.

System and method for targeted machine learning-based detection and prevention of ransomware attacks on a storage system

Techniques for performing early and adaptive IO stream sampling for ML-based optimizations in a storage system. The techniques include obtaining a sub-slice of sampled data by performing early sampling of a slice of successive operations directed to a storage object. The techniques include generating features based on the sub-slice, processing the features using an ML model, and generating a probability score based on the ML model's output. The techniques include determining that the probability score falls within an overlap range of continuous variable distributions for benign and threat classes of data. The techniques include, in response to the probability score exceeding a specified threshold, comparing a class signature of the sub-slice with a target class signature of the threat class of data to determine a similarity between the class signatures, and, in response to the similarity exceeding a predetermined similarity level, assigning a 'threat' class label to the probability score.

Early and adaptive stream sampling system and method for machine learning-based optimizations in storage systems

A method, computer program product, and computing system for processing a plurality of input/output (IO) requests associated with a plurality of storage objects of a storage system. A plurality of IO features may be generated using the plurality of IO requests including one or more of: a percentage of overwrite IO requests, a percentage of sequential read IO requests, and a percentage of sequential write IO requests. The plurality of IO features may be processed using a machine learning model. A ransomware attack may be monitored for on the storage system in real-time based upon, at least in part, the processing of the plurality of IO features using the machine learning model.

Storage and method for machine learning-based detection of ransomware attacks on a storage system

A method, computer program product, and computing system for processing a plurality of input/output (IO) requests associated with a plurality of storage objects of a storage system. A plurality of IO features are generated using the plurality of IO requests. The plurality of IO features are processed using a machine learning model. A ransomware attack on the storage system may be monitored for in real-time based upon, at least in part, the processing of the plurality of IO features using the machine learning model.

System and method for machine learning-based detection of ransomware attacks on a storage system

A method, computer program product, and computing system for processing a plurality of input/output (IO) operations on a plurality of storage objects of a storage system. The plurality of storage objects may be divided into a plurality of storage object groups based upon, at least in part, the plurality of IO operations processed on the plurality of storage objects. Each storage object group may be associated with an IO machine learning model selected from a plurality of IO machine learning models, thus defining a plurality of storage object group-specific IO machine learning models. IO performance data may be forecasted for the plurality of storage objects using the plurality of storage object group-specific IO machine learning models.

System and method for distributed input/output (IO) performance forecasting across multiple machine learning models

Techniques for performing effective noise removal for biased machine learning (ML) based optimizations in storage systems. The techniques include serving, by a storage system, an IO workload, identifying, using ML from among a plurality of storage objects subject to the IO workload, storage objects with low temperatures (e.g., cold storage objects) or likely to have low temperatures in the near future, and removing them from subsequent temperature forecasting analysis, effectively treating such cold storage objects as 'noise.' The techniques further include performing the temperature forecasting analysis on remaining ones of the plurality of storage objects such as those with high temperatures (e.g., hot storage objects). In this way, temperature forecasting or prediction is performed, using ML, in a biased fashion over a relatively narrow spectrum of storage object temperatures, thereby improving tiering and data prefetching performance, reducing memory and processing overhead, and so on.

Effective noise removal techniques for biased machine learning based optimizations in storage systems

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

A method, computer program product, and computing system for receiving a first set of input/output (IO) requests for one or more storage objects. One or more IO properties may be extracted from the first set of IO requests. Metadata may be associated with the one or more storage objects using one or more machine learning models based upon, at least in part, the one or more IO properties extracted from the first set of IO requests, thus defining storage object metadata. One or more IO processing rules may be enabled based upon, at least in part, the storage object metadata. A subsequent set of IO requests may be received. Processing of the subsequent set of IO requests on the one or more storage objects may be optimized based upon, at least in part, the storage object metadata and the one or more IO processing rules associated with the one or more storage objects.

System and method for storage optimization using machine learning- driven application and workload identification

A method, computer program product, and computing system for forecasting a future temperature for a storage object within a multi-tiered cloud storage system. A cost associated with modifying a tiering of the storage object within the multi-tiered cloud storage system is determined based upon, at least in part, the future temperature forecasted for the storage object. The storage object is tiered in the multi-tiered cloud storage system based upon, at least in part, the cost associated with modifying the tiering of the storage object and a tiering policy associated with the multi-tiered cloud storage system.

System and method for machine learning-temperature forecasting for automated tiering within a cloud storage system

A method, computer program product, and computing system for processing a plurality of input/output (IO) requests for a storage object of a storage system. A sampling interval may be determined for the plurality of IO requests for the storage object based upon, at least in part, a machine learning model processing the plurality of IO requests. The plurality of IO requests may be sampled using the determined sampling interval. The plurality of sampled IO requests may be processed using the machine learning model.

System and method for input/output (IO) request stream sampling for machine learning-based optimizations in storage systems

A data storage system includes a tuner that obtains data samples for data storage operations of workloads and calculates feature measures for a set of features of the data storage operations over aggregation intervals of an operating period. It further (1) applies a cluster analysis to the feature measures to define a set of clusters, and assigns the feature measures to the clusters, and (2) applies a classification analysis to the feature measures labelled by their clusters to identify dominating features of each cluster, and generates workload profiles for the clusters based on the dominating features, and then automatically adjusts configurable processing mechanisms (e.g., caching or tiering) based on the workload profiles and performance or efficiency goals.

Growing community of inventors

Petach Tikva, Israel

Shaul Dar