Patents by Inventor Subramaniam Venkatraman Krishnan
Subramaniam Venkatraman Krishnan has filed for patents to protect the following inventions. This listing includes patent applications that are pending as well as patents that have already been granted by the United States Patent and Trademark Office (USPTO).
-
Publication number: 20240111739Abstract: An automated tuning service is used to automatically tune, or modify, the operational parameters of a large-scale cloud infrastructure. The tuning service performs automated and fully data/model-driven configuration based from learning various real-time performance of the cloud infrastructure. Such performance is identified through monitoring various telemetric data of the cloud infrastructure. The tuning service leverages a mix of domain knowledge and principled data-science to capture the essence of our cluster dynamic behavior in a collection of descriptive machine learning (ML) models. The ML models power automated optimization procedures for parameter tuning, and inform administrators in most tactical and strategical engineering/capacity decisions (such as hardware and datacenter design, software investments, etc.).Type: ApplicationFiled: December 8, 2023Publication date: April 4, 2024Inventors: Yiwen ZHU, Subramaniam Venkatraman KRISHNAN, Konstantinos KARANASOS, Carlo CURINO, Isha TARTE, Sudhir DARBHA
-
Patent number: 11880347Abstract: An automated tuning service is used to automatically tune, or modify, the operational parameters of a large-scale cloud infrastructure. The tuning service performs automated and fully data/model-driven configuration based from learning various real-time performance of the cloud infrastructure. Such performance is identified through monitoring various telemetric data of the cloud infrastructure. The tuning service leverages a mix of domain knowledge and principled data-science to capture the essence of our cluster dynamic behavior in a collection of descriptive machine learning (ML) models. The ML models power automated optimization procedures for parameter tuning, and inform administrators in most tactical and strategical engineering/capacity decisions (such as hardware and datacenter design, software investments, etc.).Type: GrantFiled: April 2, 2021Date of Patent: January 23, 2024Assignee: Microsoft Technology Licensing, LLC.Inventors: Yiwen Zhu, Subramaniam Venkatraman Krishnan, Konstantinos Karanasos, Carlo Curino, Isha Tarte, Sudhir Darbha
-
Publication number: 20230394369Abstract: Embodiments described herein enable tracking machine learning (“ML”) model data provenance. In particular, a computing device is configured to accept ML model code that, when executed, instantiates and trains an ML model, to parse the ML model code into a workflow intermediate representation (WIR), to semantically annotate the WIR to provide an annotated WIR, and to identify, based on the annotated WIR and ML API corresponding to the ML model code, data from at least one data source that is relied upon by the ML model code when training the ML model. A WIR may be generated from an abstract syntax tree (AST) based on the ML model code, generating provenance relationships (PRs) based at least in part on relationships between nodes of the AST, wherein a PR comprises one or more input variables, an operation, a caller, and one or more output variables.Type: ApplicationFiled: August 21, 2023Publication date: December 7, 2023Inventors: Avrilia FLORATOU, Ashvin AGRAWAL, MohammadHossein NAMAKI, Subramaniam Venkatraman KRISHNAN, Fotios PSALLIDAS, Yinghui WU
-
Publication number: 20230385649Abstract: Linguistic schema mapping via semi-supervised learning is used to map a customer schema to a particular industry-specific schema (ISS). The customer schema is received and a corresponding ISS is identified. An attribute in the customer schema is selected for labeling. Candidate pairs are generated that include the first attribute and one or more second attributes which may describe the first attribute. A featurizer determines similarities between the first attribute and second attribute in each generated pair, one or more suggested labels are generated by a machine learning (ML) model, and one of the suggested labels is applied to the first attribute.Type: ApplicationFiled: May 28, 2022Publication date: November 30, 2023Inventors: Avrilia FLORATOU, Joyce Yu CAHOON, Subramaniam Venkatraman KRISHNAN, Andreas C. MUELLER, Dalitso Hansini BANDA, Fotis PSALLIDAS, Jignesh PATEL, Yunjia ZHANG
-
Patent number: 11775862Abstract: A system enables tracking machine learning (“ML”) model data provenance. In particular, a computing device is configured to accept ML model code that, when executed, instantiates and trains an ML model, to parse the ML model code into a workflow intermediate representation (WIR), to semantically annotate the WIR to provide an annotated WIR, and to identify, based on the annotated WIR and ML API corresponding to the ML model code, data from at least one data source that is relied upon by the ML model code when training the ML model. A WIR may be generated from an abstract syntax tree (AST) based on the ML model code, generating provenance relationships (PRs) based at least in part on relationships between nodes of the AST, wherein a PR comprises one or more input variables, an operation, a caller, and one or more output variables.Type: GrantFiled: January 14, 2020Date of Patent: October 3, 2023Assignee: MICROSOFT TECHNOLOGY LICENSING, LLCInventors: Avrilia Floratou, Ashvin Agrawal, MohammadHossein Namaki, Subramaniam Venkatraman Krishnan, Fotios Psallidas, Yinghui Wu
-
Publication number: 20230029888Abstract: Methods, systems, apparatuses, and computer-readable storage mediums described herein are directed to determining and recommending an optimal compute resource configuration for a cloud-based resource (e.g., a server, a virtual machine, etc.) for migrating a customer to the cloud. The embodiments described herein utilize a statistically robust approach that makes recommendations that are more flexible (elastic) and account for the full distribution of the amount of resource usage. Such an approach is utilized to develop a personalized rank of relevant recommendations to a customer. To determine which compute resource configuration to recommend to the customer, the customer’s usage profile is matched to a set of customers that have already migrated to the cloud. The compute resource configuration that reaches the performance most similar to the performance of the configurations utilized by customers in the matched set is recommended to the user.Type: ApplicationFiled: December 20, 2021Publication date: February 2, 2023Inventors: Wenjing WANG, Joyce Yu CAHOON, Yiwen ZHU, Ya LIN, Subramaniam Venkatraman KRISHNAN, Neetu SINGH, Raymond TRUONG, XingYu LIU, Maria Alexandra CIORTEA, Sreraman NARASIMHAN, Pratyush RAWAT, Haitao SONG
-
Publication number: 20220164327Abstract: An automated tuning service is used to automatically tune, or modify, the operational parameters of a large-scale cloud infrastructure. The tuning service performs automated and fully data/model-driven configuration based from learning various real-time performance of the cloud infrastructure. Such performance is identified through monitoring various telemetric data of the cloud infrastructure. The tuning service leverages a mix of domain knowledge and principled data-science to capture the essence of our cluster dynamic behavior in a collection of descriptive machine learning (ML) models. The ML models power automated optimization procedures for parameter tuning, and inform administrators in most tactical and strategical engineering/capacity decisions (such as hardware and datacenter design, software investments, etc.).Type: ApplicationFiled: April 2, 2021Publication date: May 26, 2022Inventors: Yiwen ZHU, Subramaniam Venkatraman KRISHNAN, Konstantinos KARANASOS, Carlo CURINO, Isha TARTE, Sudhir Darbha
-
Publication number: 20210216905Abstract: Embodiments described herein enable tracking machine learning (“ML”) model data provenance. In particular, a computing device is configured to accept ML model code that, when executed, instantiates and trains an ML model, to parse the ML model code into a workflow intermediate representation (WIR), to semantically annotate the WIR to provide an annotated WIR, and to identify, based on the annotated WIR and ML API corresponding to the ML model code, data from at least one data source that is relied upon by the ML model code when training the ML model. A WIR may be generated from an abstract syntax tree (AST) based on the ML model code, generating provenance relationships (PRs) based at least in part on relationships between nodes of the AST, wherein a PR comprises one or more input variables, an operation, a caller, and one or more output variables.Type: ApplicationFiled: January 14, 2020Publication date: July 15, 2021Inventors: Avrilia Floratou, Ashvin Agrawal, MohammadHossein Namaki, Subramaniam Venkatraman Krishnan, Fotios Psallidas, Yinghui Wu
-
Publication number: 20210124739Abstract: The description relates to executing an inference query relative to a database management system, such as a relational database management system. In one example a trained machine learning model can be stored within the database management system. An inference query can be received that applies the trained machine learning model on data local to the database management system. Analysis can be performed on the inference query and the trained machine learning model to generate a unified intermediate representation of the inference query and the trained model. Cross optimization can be performed on the unified intermediate representation. Based upon the cross-optimization, a first portion of the unified intermediate representation to be executed by a database engine of the database management system can be determined, and, a second portion of the unified intermediate representation to be executed by a machine learning runtime can be determined.Type: ApplicationFiled: August 11, 2020Publication date: April 29, 2021Applicant: Microsoft Technology Licensing, LLCInventors: Konstantinos KARANASOS, Matteo INTERLANDI, Fotios PSALLIDAS, Rathijit SEN, Kwanghyun PARK, Ivan POPIVANOV, Subramaniam VENKATRAMAN KRISHNAN, Markus WEIMER, Yuan YU, Raghunath RAMAKRISHNAN, Carlo Aldo CURINO, Doris Suiyi XIN, Karla Jean SAUR
-
Publication number: 20210089532Abstract: The cloud-based query workload optimization system disclosed herein the cloud-based query workloads optimization system receives query logs from various query engines to a cloud data service, extracts various query entities from the query logs, parses query entities to generate a set of common workload features, generates intermediate representations of the query workloads, wherein the intermediate representations are agnostic to the language of the plurality of the queries, identifies a plurality of workload patterns based on the intermediate representations of the query workloads, categorizes the workloads in one or more workload type categories based on the workload patterns and the workload features, and selects an optimization scheme based on the category of workload pattern.Type: ApplicationFiled: September 25, 2019Publication date: March 25, 2021Inventors: Hiren S. PATEL, Rathijit SEN, Zhicheng YIN, Shi QIAO, Abhishek ROY, Alekh JINDAL, Subramaniam Venkatraman KRISHNAN, Carlo Aldo CURINO
-
Patent number: 10936367Abstract: Described herein is a system and method for ranking and/or taking an action regarding execution of jobs of a shared computing cluster based upon predicted user impact. Information regarding previous executions of a plurality of jobs is obtained, for example, from job execution log(s). Data dependencies of the plurality of jobs are determined. Job impact of each of the plurality of jobs as a function of the determined data dependencies is calculated. User impact of each of the plurality of jobs as a function of the determined data dependencies, the calculated job impact, and time is calculated. The plurality of jobs are ranked in accordance with the calculated user impact. An action is taken in accordance with the ranking of the plurality of jobs. The action can include automatic scheduling of the jobs and/or providing information regarding the rankings to a user.Type: GrantFiled: January 22, 2019Date of Patent: March 2, 2021Assignee: Microsoft Technology Licensing, LLCInventors: Carlo Aldo Curino, Konstantinos Karanasos, Subramaniam Venkatraman Krishnan, Christopher William Douglas, Sriram S Rao, Andrew F Chung
-
Publication number: 20200133726Abstract: Described herein is a system and method for ranking and/or taking an action regarding execution of jobs of a shared computing cluster based upon predicted user impact. Information regarding previous executions of a plurality of jobs is obtained, for example, from job execution log(s). Data dependencies of the plurality of jobs are determined. Job impact of each of the plurality of jobs as a function of the determined data dependencies is calculated. User impact of each of the plurality of jobs as a function of the determined data dependencies, the calculated job impact, and time is calculated. The plurality of jobs are ranked in accordance with the calculated user impact. An action is taken in accordance with the ranking of the plurality of jobs. The action can include automatic scheduling of the jobs and/or providing information regarding the rankings to a user.Type: ApplicationFiled: January 22, 2019Publication date: April 30, 2020Applicant: Microsoft Technology Licensing, LLCInventors: Carlo Aldo CURINO, Konstantinos KARANASOS, Subramaniam VENKATRAMAN KRISHNAN, Christopher William DOUGLAS, Sriram S. RAO, Andrew F. CHUNG
-
Patent number: 9876878Abstract: Embodiments are directed to progressively migrating source computer nodes where the source computer nodes perform a computer-implemented service. In one embodiment, a computer system determines that execution of the performed service is to be migrated from the source computer nodes to target computer nodes. The computer system groups the source computer nodes into multiple source subgroups, where each source subgroup includes at least one source computer node. The computer system then schedules creation of target subgroups of target nodes. These target subgroups include at least one source computer node and, themselves, correspond to a source subgroup. The computer system activates a first target subgroup corresponding to a first source subgroup, and deactivates the first source subgroup. In this manner, the first target subgroup replaces the first source subgroup. Still further, the target subgroups are scheduled to be created only after the first source subgroup has been deactivated.Type: GrantFiled: January 20, 2017Date of Patent: January 23, 2018Assignee: Microsoft Technology Licensing, LLCInventors: Kishore Chaliparambil, Carlo Curino, Kannababu Karanam, Subramaniam Venkatraman Krishnan, Christopher William Douglas, Sriram Rao, Mostafa Elhemali, Chuan Liu
-
Patent number: 9699250Abstract: A method and system for building an elastic cloud web server farm. The method includes registering a web application on a serving cloud and copying the web application to a distributed store. A load of the web application is specified, and a plurality of nodes is added for the web application based on the load. A web server corresponding to a node of the plurality of nodes is then initialized. A web request is received from a user and a web server is selected from a list of available web servers to process the web request. The web request is further transmitted to the web server. A web response, based on the web request, is transmitted back to the user. The system includes a central registry, a distributed store, a process coordinator, one or more web servers, and a router.Type: GrantFiled: January 12, 2015Date of Patent: July 4, 2017Assignee: EXCALIBUR IP, LLCInventors: Subramaniam Venkatraman Krishnan, Amit Jaiswal, Ravikaran Meka, Jean Christophe Counio, Alejandro Abdelnur, Ruchir Rajendra Shah
-
Publication number: 20170134526Abstract: Embodiments are directed to progressively migrating source computer nodes where the source computer nodes perform a computer-implemented service. In one embodiment, a computer system determines that execution of the performed service is to be migrated from the source computer nodes to target computer nodes. The computer system groups the source computer nodes into multiple source subgroups, where each source subgroup includes at least one source computer node. The computer system then schedules creation of target subgroups of target nodes. These target subgroups include at least one source computer node and, themselves, correspond to a source subgroup. The computer system activates a first target subgroup corresponding to a first source subgroup, and deactivates the first source subgroup. In this manner, the first target subgroup replaces the first source subgroup. Still further, the target subgroups are scheduled to be created only after the first source subgroup has been deactivated.Type: ApplicationFiled: January 20, 2017Publication date: May 11, 2017Inventors: Kishore Chaliparambil, Carlo Curino, Kannababu Karanam, Subramaniam Venkatraman Krishnan, Christopher William Douglas, Sriram Rao, Mostafa Elhemali, Chuan Liu
-
Patent number: 9578091Abstract: Embodiments are directed to progressively migrating source computer nodes where the source computer nodes perform a computer-implemented service. In one embodiment, a computer system determines that execution of the performed service is to be migrated from the source computer nodes to target computer nodes. The computer system groups the source computer nodes into multiple source subgroups, where each source subgroup includes at least one source computer node. The computer system then schedules creation of target subgroups of target nodes. These target subgroups include at least one source computer node and, themselves, correspond to a source subgroup. The computer system activates a first target subgroup corresponding to a first source subgroup, and deactivates the first source subgroup. In this manner, the first target subgroup replaces the first source subgroup. Still further, the target subgroups are scheduled to be created only after the first source subgroup has been deactivated.Type: GrantFiled: December 30, 2013Date of Patent: February 21, 2017Assignee: Microsoft Technology Licensing, LLCInventors: Kishore Chaliparambil, Carlo Curino, Kannababu Karanam, Subramaniam Venkatraman Krishnan, Christopher William Douglas, Sriram Rao, Mostafa Elhemali, Chuan Liu
-
Patent number: 9558457Abstract: A method and system for automatically identifying optimal meeting locations. The method includes receiving a plurality of meeting parameters associated with one or more participants. The method also includes identifying a list of optimal meeting locations relevant to one or more of the plurality of meeting parameters. The method further includes ranking the list of optimal meeting locations. Further, the method includes enabling a user to select an optimal meeting location from the list of optimal meeting locations. The system includes one or more electronic devices and a user electronic device. The user electronic device includes a communication interface, a memory, and a processor.Type: GrantFiled: July 26, 2011Date of Patent: January 31, 2017Assignee: EXCALIBUR IP, LLCInventors: Deepak Kumar V, Subramaniam Venkatraman Krishnan, Ashvin Agrawal
-
Patent number: 9311628Abstract: An appointment having an associated appointment location and a reminder time is received. The method also includes tracking a current location and a travel time, the travel time comprising an estimated amount of time for travel from the current location to the appointment location. Further, the method includes adjusting the reminder time to accommodate the travel time. Furthermore, the method includes activating an event reminder in accordance with the adjusted reminder time.Type: GrantFiled: December 22, 2010Date of Patent: April 12, 2016Assignee: Yahoo! Inc.Inventors: Ashvin Agrawal, Subramaniam Venkatraman Krishnan
-
Publication number: 20150188989Abstract: Embodiments are directed to progressively migrating source computer nodes where the source computer nodes perform a computer-implemented service. In one embodiment, a computer system determines that execution of the performed service is to be migrated from the source computer nodes to target computer nodes. The computer system groups the source computer nodes into multiple source subgroups, where each source subgroup includes at least one source computer node. The computer system then schedules creation of target subgroups of target nodes. These target subgroups include at least one source computer node and, themselves, correspond to a source subgroup. The computer system activates a first target subgroup corresponding to a first source subgroup, and deactivates the first source subgroup. In this manner, the first target subgroup replaces the first source subgroup. Still further, the target subgroups are scheduled to be created only after the first source subgroup has been deactivated.Type: ApplicationFiled: December 30, 2013Publication date: July 2, 2015Applicant: Microsoft CorporationInventors: Kishore Chaliparambil, Carlo Curino, Kannababu Karanam, Subramaniam Venkatraman Krishnan, Christopher William Douglas, Sriram Rao, Mostafa Elhemali, Chuan Liu
-
Publication number: 20150127725Abstract: A method and system for building an elastic cloud web server farm. The method includes registering a web application on a serving cloud and copying the web application to a distributed store. A load of the web application is specified, and a plurality of nodes is added for the web application based on the load. A web server corresponding to a node of the plurality of nodes is then initialized. A web request is received from a user and a web server is selected from a list of available web servers to process the web request. The web request is further transmitted to the web server. A web response, based on the web request, is transmitted back to the user. The system includes a central registry, a distributed store, a process coordinator, one or more web servers, and a router.Type: ApplicationFiled: January 12, 2015Publication date: May 7, 2015Inventors: Subramaniam Venkatraman Krishnan, Amit Jaiswal, Ravikaran Meka, Jean Christophe Counio, Alejandro Abdelnur, Ruchir Rajendra Shah