International Business Machines Corporation

Darrell Christopher Reimer

Vasanth Bala

Todd W Mummert

Glenn S Ammons

Bowen L Alpern

Mark N Wegman

Xiaolan Zhang

Arun Thomas

Evelyn Duesterwald

Canturk Isci

Stephen Fink

Sastry Duri

Matthew R Arnold

Bard Bloom

Ian D Simmonds

Johannes Christiaan Laffra

Alan A Donovan

Sameep Mehta

Ramya Raghavendra

Kavitha Srinivas

Kalapriya Kannan

Kush Raj Varshney

John Thomas Richards

Ea-Ee Jan

Aleksandra Mojsilovic

Giovanni Pacifici

Rachel K E Bellamy

Manish Kesarwani

Robert Leslie Yates

Lakshminarayanan Renganarayana

Jian Yin

Kaoutar El Maghraoui

Michael Desmond

Harold Leon Ossher

Michael James Hind

Mukund Raghavachari

Balachandar Rajaraman

Robert D Johnson

James Arthur Lindeman

Stephanie Houde

David John Piorkowski

Benjamin W Herta

Shaikh Shahriar Quader

Jason Tsay

Punleuk Oum

Niyu Ge

Debashish Saha

Archit Verma

Gaodan Fang

Mona Nashaat Ali Elmowafy

Mona Attariyan

A method of using a computing device to manage a lifecycle of machine learning models includes receiving, by a computing device, multiple pre-defined machine learning lifecycle tasks. The computing device manages executing a management-layer software layer for the multiple pre-defined machine learning lifecycle tasks. The computing device further generates and updates a machine learning pipeline using the management-layer software layer.

Configurable and scalable machine learning model lifecycle operator

Noisy labeled and unlabeled datapoint detection and rectification in a training dataset for machine-learning is facilitated by a processor(s) obtaining a training dataset for use in training a machine-learning model. The processor(s) applies ensemble machine-learning and a generative model to the training dataset to detect noisy labeled datapoints in the training dataset, and create a clean dataset with preliminary labels added for any unlabeled datapoints in the training dataset. Data-driven active learning and the clean dataset are used by the processor(s) to facilitate generating an active-learned dataset with true labels added for one or more selected datapoints of a datapoint pool including the detected noisy labeled datapoints and the unlabeled datapoints of the training dataset. The machine-learning model is trained by the processor(s) using, at least in part, the clean dataset and the active-learned dataset.

Data-analysis-based, noisy labeled and unlabeled datapoint detection and rectification for machine-learning

The generating of actionable recommendations for tuning model metrics of an Artificial Intelligence (AI) system includes partitioning a key performance indicator (KPI) range associated with a target system into a plurality of buckets. Log data including at least one KPI of the target system and one or more AI model metrics is partitioned and distributed across the plurality of buckets. For each bucket, an aggregate value of the one or more AI model metrics across the log data is computed and weighted according to the volume of log data in that bucket. A correlation factor between the aggregate value and a representative KPI value for each bucket is determined. A model tuning recommendation to increase ranking of the AI model metrics according to the determined correlation factor is provided to an output device and/or to the AI system for updating the one or more AI model.

Generating model insights by progressive partitioning of log data across a set of performance indicators

A method for automatically generating documentation for an artificial intelligence model includes receiving, by a computing device, an artificial intelligence model. The computing device accesses a model facts policy that indicates data to be collected for artificial intelligence models. The computing device collects artificial intelligence model facts regarding the artificial intelligence model according to the model facts policy. The computing device accesses a factsheet template. The factsheet template provides a schema for an artificial intelligence model factsheet for the artificial intelligence model. The computing device populates the artificial intelligence model factsheet using the factsheet template with the artificial intelligence model facts related to the artificial intelligence model.

Generation and management of an artificial intelligence (AI) model documentation throughout its life cycle

A system and method are provided for communicating information in a data structure between applications. According to the method, a description of a data structure is sent from a first application to a second application, and there is received from the second application an identification of at least one portion of the data structure that is requested by the second application. The first application marshals a subset of the data structure consisting of the at least one portion that was identified, and there is sent from the first application to the second application the marshalled subset of the data structure.

Protocol for communication of data structures

Aspects of the present invention provide a solution for passively monitoring a computer system. In an embodiment, a read-only target is created based on the configuration of the virtual server disk used by the virtual server to access the storage associated therewith. This read-only target can be used to expose the virtual server disk for access by a monitoring crawler. Using the read-only target, a monitoring crawler can acquire a set of features regarding the virtual server from the virtual server disk. One or more of this set of features can be analyzed to perform passive monitoring of the virtual server.

Passive monitoring of virtual systems using agent-less, near-real-time indexing

Aspects of the present invention provide a solution for passively monitoring a computer system. In an embodiment, an indexing agent operates on a physical server having a plurality of virtual servers. The indexing agent has a framework having a plurality of indexing layers. Each of these indexing layers is adapted to index a virtual server along a different abstraction level to extract abstraction level features indicative of changes in the virtual server. Further, each of the indexing layers has an application programming interface, which is adapted to receive one or more external indexing modules to be plugged into the framework. Each indexing module is adapted to provide instructions to the indexing agent while the indexing layer to which it is plugged is indexing the associated abstraction level of the virtual machine.

Passive monitoring of virtual systems using extensible indexing

Aspects of the present invention provide a solution for creating a snapshot of a virtual server. In an embodiment, a first write storage is created in response to a snapshot request. For each write from the virtual server to primary storage that is the first write of that particular data block since the snapshot request, the data that resided in the data block previous to the write is copied from the primary storage to the first write storage.

Minimally disruptive virtual machine snapshots

A composite software stack is distributed as a virtual appliance. A new virtual appliance is published by obtaining a parent virtual appliance; configuring the new virtual appliance; obtaining file level differences between the parent virtual appliance and the new virtual appliance; and publishing the new virtual appliance as a patch comprising the file level differences relative to the parent virtual appliance. The file level differences comprise additions, deletions and/or modifications to the parent virtual appliance. An update to a given virtual appliance can be published as a fix pack, wherein the fix pack comprises file level differences relative to the given virtual appliance. An end user obtains a desired virtual appliance by selecting the desired virtual appliance; obtaining metadata for the desired virtual appliance comprising an identifier of a patch file and a parent virtual appliance; and applying the patch file to the parent virtual appliance to get a modified virtual appliance.

Method and apparatus for distributing a composite software stack as a virtual machine image

Aspects of the present invention provide a tool for analyzing and remediating an update-related failure. In an embodiment, a failure state of a computer system that has been arrived at as a result of an update is captured. A semantic diff that includes the difference between the failure state and at least one of an original state or a completion state is then computed. This semantic diff is transformed into a feature vector format. Then the transformed semantic diff is analyzed to determine a remediation for the update. Failure and/or resolution signatures can be constructed using the semantic diff and contextual data, and these signatures can be used in comparison and analysis of failures and resolutions.

Systematic failure remediation

A method for managing a virtual machine image includes receiving a request to change a package configuration of a machine, processing an image action received in the request, wherein the processing the image action received includes, opening the image action with associated action inputs, comparing the opened image action inputs with inputs associated with action instances in a database, determining whether the compared opened image action inputs match the inputs associated with action instances, retrieving from the database, image difference data associated with the image action responsive to determining that the compared opened image action inputs match the inputs associated with action instances in the database, and applying the image difference data to an image to transform the image, determining whether each image action in the request has been processed and processing a second image action responsive to determining that each image action has not been processed.

Scalable package management for virtual-machine images

A novel method is disclosed for capturing an installed state of a conventional application and converting the captured state into a virtual application. The novel method starts with a bare machine in a known state, preferably soon after the OS was installed. Installation scripts are used to install one or more software applications along with required components and dependencies. Other artifacts can be added and configured such as files, trees, directories, entries, data, values, among others. These also may include updates to various system databases, such as the Windows registry in which certain metadata is stored. The installed applications are tested and verified to work as desired. Undesired artifacts can be deleted manually or removed by the OS. The state of the virtual machine with the installed applications is captured. The installed applications can be tested on the frozen virtual machine. If the user determines that the installed applications are working properly, a virtual machine image is prepared and then written out. The resulting output is a new populated virtual application container that encapsulates the applications installed and configured by the user. If the user determines that one or more installed applications do not work properly during testing, then the frozen virtual machine can be unfrozen and any deficiencies corrected. The virtual machine image is stored as a launch document and a set of shards. The virtual machine image represents the state of the virtual machine along with the installed applications at the time the preparation virtual machine was frozen.

Creating an application virtual machine image by isolating installation artifacts in shadow area

Methods and apparatus are provided for distributing a composite software stack as a virtual appliance. A new virtual appliance is published by obtaining a parent virtual appliance; configuring the new virtual appliance; obtaining file level differences between the parent virtual appliance and the new virtual appliance; and publishing the new virtual appliance as a patch comprising the file level differences relative to the parent virtual appliance. The file level differences comprise one or more of additions, deletions and modifications to the parent virtual appliance. An update to a given virtual appliance can be published as a fix pack, wherein the fix pack comprises file level differences relative to the given virtual appliance. An end user obtains a desired virtual appliance is obtained (for example, from a library of virtual appliances) by selecting the desired virtual appliance; obtaining metadata for the desired virtual appliance comprising an identifier of a patch file and a parent virtual appliance; and applying the patch file to the parent virtual appliance to get a modified virtual appliance.

A method for updating an application on a host system includes steps or acts of: installing an application on the host system; installing a virtual machine monitor on the host system, installing a first virtual container on the host system, wherein the first virtual container comprises at least one update to the installed application; and instantiating the first virtual container under the virtual machine monitor in a mode wherein the host system can be accessed but not modified and wherein instantiating the first virtual container includes updating the installed application in the first virtual computer and running the updated application in the first virtual container. The method may also include a step of confirming that the updated application runs properly. Optionally, the virtual container may be devirtualized.

Method for delivering, testing, and applying software patches or other changes to a conventionally installed application in virtual application containers

Growing community of inventors

Tarrytown, NY, United States of America

Darrell Christopher Reimer