Microsoft Technology Licensing, LLC

Qingwei Lin

Fan Li

Jiang Tao Li

Jian-guang Lou

Yusuo Hu

Yong Xu

Si Qin

Thomas Moscibroda

Karthikeyan Subramanian

Shandan Zhou

Kaixin Sui

Ajay Mani

Saurabh Agarwal

Nisarg Tarakkumar Sheth

Sorin Iftimie

Wanghai Gu

Sandeep Ramji

Paul Naveen Selvaraj

Xu Zhang

Dongmei Zhang

Yingnong Dang

Zhangwei Xu

Vivek Gupta

Chetan Bansal

John Lawrence Miller

Kamaljit S Bath

Li Yang

Bo Qiao

Eli Cortez

Murali Mohan Chintalapati

Jian-Guang Lou

Inigo Goiri Presa

Jian Zhang

Yu Kang

Rakesh Namineni

Eli Cortez Custodio Vilarinho

Ken Hsieh

Liqun Li

Xiaofeng Gao

Shanti Kemburu

Jue Zhang

Felipe Vieira Frujeri

Yandan Wang

Mahmoud Sayed

Jeffrey Ding He

Pooja Rani

Jimmy Wong

Christopher Cowdery

Robert Edward Minneker

Alexander David Fischer

Youjiang Wu

Preston Tapley Stephenson

Randolph Yao

Soumya Ram

Sidhanth M Panjwani

Camille Jean Couturier

Fangkai Yang

Bowen Pang

Namrata Jain

Mohit Verma

Rulei Yu

Pu Zhao

Jian Sun

Pochian Lee

Julien Hoachuck

A method for dynamically adjusting a number of virtual machines for a workload, includes: receiving a probability indicator for each of a plurality of N sequential stages, where N is a natural number greater than 1, of a likelihood that a virtual machine assigned to a workload will be evicted during the N sequential stages; predicting a target number of virtual machines to configure in a current stage for a subsequent stage from among the plurality of N sequential stages based on the probability indicator, a target capacity for the workload, and a current price for maintaining a virtual machine; and configuring a number of virtual machines for the workload during the current stage based on the target number to be loaded for the workload for the subsequent stage.

System and method of dynamically adjusting virtual machines for a workload

Systems and techniques for multi-phase cloud service node error prediction are described herein. A set of spatial metrics and a set of temporal metrics may be obtained for node devices in a cloud computing platform. The node devices may be evaluated using a spatial machine learning model and a temporal machine learning model to create a spatial output and a temporal output. One or more potentially faulty nodes may be determined based on an evaluation of the spatial output and the temporal output using a ranking model. The one or more potentially faulty nodes may be a subset of the node devices. One or more migration source nodes may be identified from one or more potentially faulty nodes. The one or more migration source nodes may be identified by minimization of a cost of false positive and false negative node detection.

Multi-phase cloud service node error prediction based on minimization function with cost ratio and false positive detection

Systems and techniques for multi-factor cloud service storage device error prediction are described herein. A set of storage device metrics and a set of computing system metrics may be obtained. A feature set may be generated using the set of storage device metrics and the set of computing system metrics. Members of the feature set may be validated by evaluating a validation training dataset using the members of the feature set. A modified feature set may be created based on the validation. A storage device failure model may be created using the modified feature set. A storage device rating range may be determined by minimizing a cost of misclassification of a storage device. A set of storage devices to be labeled may be identified as having a high probability of failure.

Multi-factor cloud service storage device error prediction

Methods and systems for detecting systemwide service issues by using anomaly localization. In an example, a method includes receiving time-series monitoring data for multiple services, the time-series monitoring data including multiple dimensions and an error metric; for the monitoring data from each service, evaluating scopes within the monitoring data based on an objective function for a time-series of the error metric to identify at least one anomalous scope, each scope including at least one dimension and a value for the dimension; based on evaluating the scopes, generating a ranked list of scopes for each service based on objective function scores for the scopes; correlating the ranked lists of scopes across the multiple services to identify a cross-service anomaly; and generating an alert for the services based on the cross-service anomaly, the alert indicating at least one scope as a potential root cause for the cross-service anomaly.

Detecting systemwide service issues by using anomaly localization

According to implementations of the subject matter described herein, there is provided a solution of providing a health index of a service. In this solution, a plurality of incident information sets associated with a plurality of services are obtained. The plurality of services are provisioned in a computing environment. An incident information set indicates at least one incident reported during operation of a service. Respective health indices are determined for the plurality of services based on respective ones of the plurality of incident information sets and a health classification policy. The respective health indices indicate respective health statuses of the plurality of services and being determined from a same health index range. Through unified use of incident information, the determined health indices can indicate universal and consistent health statuses for different services.

Health index of a service

The present disclosure relates to systems, methods, and computer readable media for predicting expansion failures and implementing defragmentation instructions based on the predicted expansion failures and other signals. For example, systems disclosed herein may apply a failure prediction model to determine an expansion failure prediction associated with an estimated likelihood that deployment failures will occur on a node cluster. The systems disclosed herein may further generate defragmentation instructions indicating a severity level that a defragmentation engine may execute on a cluster level to prevent expansion failures while minimizing negative customer impacts. By uniquely generating defragmentation instructions for each node cluster, a cloud computing system can minimize expansion failures, increase resource capacity, reduce costs, and provide access to reliable services to customers.

Predicting expansion failures and defragmenting cluster resources

The present disclosure relates to systems, methods, and computer readable media for predicting deployment growth on one or more node clusters and selectively permitting deployment requests on a per cluster basis. For example, systems disclosed herein may apply tenant growth prediction system trained to output a deployment growth classification indicative of a predicted growth of deployments on a node cluster. The system disclosed herein may further utilize the deployment growth classification to determine whether a deployment request may be permitted while maintaining a sufficiently sized capacity buffer to avoid deployment failures for existing deployments previously implemented on the node cluster. By selectively permitting or denying deployments based on a variety of factors, the systems described herein can more efficiently utilize cluster resources on a per-cluster basis without causing a significant increase in deployment failures for existing customers.

Allocating cloud resources in accordance with predicted deployment growth

In implementations of the subject matter described herein, a solution for query processing is provided. In this solution, data subsets are pre-stored for example in a fast access storage device for data analysis, each including data entries corresponding to one or more dimensions. If two or more data subsets are needed to cover target dimensions corresponding to query items in a received query, instead of turning to analyze a source data set that is not stored, the query is decomposed into subqueries. By means of the decomposing, the target dimension(s) corresponding to the query item(s) in each subquery can be covered by a single data subset. The data subset is analyzed for each subquery and a query result for the query is determined based on analysis results of the subqueries. In such way, the query result for the query can obtained in a fast manner from the available data subsets.

Query processing in data analysis

A system may include a node historical state data store having historical node state data, including a metric that represents a health status or an attribute of a node during a period of time prior to a node failure. A node failure prediction algorithm creation platform may generate a machine learning trained node failure prediction algorithm. An active node data store may contain information about computing nodes in a cloud computing environment, including, for each node, a metric that represents a health status or an attribute of that node over time. A virtual machine assignment platform may then execute the node failure prediction algorithm to calculate a node failure probability score for each computing node based on the information in the active node data store. As a result, a virtual machine may be assigned to a selected computing node based at least in part on node failure probability scores.

Computing node failure and health prediction for cloud-based data center

Data for performing help document animated visualization is obtained by generating operation records from a text-based help document of an application. Each of the operation records may include data for animating an operation action that is performed on a user interface (UI) element of the application. The help document is further enhanced to include controls that load the operation records. The enhanced help documents and the operation records are distributed for use.

Help document animated visualization

Described herein are techniques for automatically batching GUI-based (Graphical User Interface) tasks. The described techniques include automatically determining whether a user is performing batchable tasks in a GUI-based environment. Once detected, the described techniques include predicting the next tasks of a batch based upon those detected batchable tasks. With the described techniques, the user may be asked to verify and/or correct the predicted next tasks. Furthermore, the described techniques may include performing a batch and doing so without user interaction.

Automatic batching of GUI-based tasks

Techniques described herein enable peers to determine each peer's NAT type much more efficiently and quickly than when compared with existing techniques. To do so, a peer simultaneously sends multiple test messages to a server. The peer then waits to either receive a response for each of the multiple test messages or may store an indication that no response has been received after a predetermined timeout period. The peer then analyzes the received responses and/or the stored timeout indications to determine the peer's NAT type or to determine that the peer is operating free from concealment by a NAT/firewall device. By simultaneously sending the multiple test messages, the peer may determine the NAT type within a maximum time defined by the predetermined timeout period or a roundtrip time period that is required for communication between the peer and the server. As such, the tools allow for efficient NAT-type detection.

Network address translators (NAT) type detection techniques

Growing community of inventors

Beijing, China

Qingwei Lin