Oracle International Corporation

Angelo Pruscino

Michael Zoll

Wilson Wai Chan

Wilson Wai Shun Chan

Tak Fung Wang

Sashikanth Chandrasekaran

Tolga Yurek

Yaser I Suleiman

Roger J Bamford

Thomas Herter

Stefan Heinrich Roesch

Subhransu Basu

Wataru Miyoshi

Thomas Breidt

Klaus Thielen

Sahil Kumar

Wolfgang Lohwasser

Frederick S Glover

Diane Lebel

Eugene Reddy Ho

James Aubrey Williams

William Bridge

Sameer Arun Joshi

David Brower

John C Hsu

Richard L Long

Cheng-Lu Hsu

Deepti Srivastava

Thomas V Sepez

Thomas J Engle

Patrick Ritto

Alok Kumar Srivastava

Carol L Colrain

William Havinden Bridge, Jr

Vinay H Srihari

Neil MacNaughton

Sunil Kumar

Fulu Li

Yu Kin Ho

Mohsin Beg

Timothy Mark

Sujeet Kumar

Donald Allan Graves, Jr

David Akers

Vikram Rai

Joy Mukherjee

Anuj Tripathi

Macharapu Sai Prasanth

Ahmed S Abbas

WeiHsin Gu

Fred S Glover

James Beltz

Brian Tsao

Stefan Pommerenk

Beverly Zane

William Jr Bridge

Neil Macnaughton

Thomas Sepez

Described is an approach that provides an adaptive solution to missing data for machine learning systems. A gradient solution is provided that is attentive to imputation needs at each of several missingness levels. This multilevel approach treats data missingness at any of multiple severity levels while utilizing, as much as possible, the actual observed data.

Method and system for adaptively imputing sparse and missing data for predictive models

Described is an improved approach to remove data outliers by filtering out data correlated to detrimental events within a system. One or more detrimental even conditions are defined to identify and handle abnormal transient states from collected data for a monitored system.

Method and system for adaptively removing outliers from data used in training of predictive models

Described is an approach for performing context-aware prognoses in machine learning systems. The approach harnesses streams of detailed data collected from a monitored target to create a context, in parallel to ongoing model operations, for the model outcomes. The context is then probed to identify the particular elements associated with the model findings.

Method and system for performing context-aware prognoses for health analysis of monitored systems

Described is an improved approach to implement selection of training data for machine learning, by presenting a designated set of specific data indicators where these data indicators correspond to metrics that end users are familiar with and are easily understood by ordinary users and DBAs within their knowledge domain. Selection of these indicators would correlate automatically to selection of a corresponding set of other metrics/signals that are less understandable to an ordinary user. Additional analysis of the selected data can then be performed to identify and correct any statistical problems with the selected training data.

Method and system for cleansing training data for predictive models

A method, system, and computer program product for generating database cluster health alerts using machine learning. A first database cluster known to be operating normally is measured and modeled using machine learning techniques. A second database cluster is measured and compared to the learned model. More specifically, the method collects a first set of empirically-measured variables of a first database cluster, and using the first set of empirically-measured variables a mathematical behavior predictor model is generated. Then, after collecting a second set of empirically-measured variables of a second database cluster over a plurality of second time periods, the mathematical behavior predictor model classifies the observed behavior. The classified behavior might be deemed to be normal behavior, or some form of abnormal behavior. The method forms and report alerts when the classification deemed to be anomalous behavior, or fault behavior. A Bayesian belief network predicts the likelihood of continued anomalous behavior.

Generating database cluster health alerts using machine learning

Techniques are provided for processing file system requests using a super cluster of clusters of nodes. Mirror file systems for processing the requests are exported through multiple clusters in the super cluster. A cluster may be assigned to an active or passive role for processing file system requests for a set of mirror file systems. A super cluster bundle, or mapping between a cluster in the super cluster and a file system resource on the set of mirror file systems, is created to process the file system requests. The super cluster bundle represents an amount of work assigned to the cluster. A super cluster bundle is reassigned from one cluster to another in response to a failover, or in response to a load balancing determination.

Highly available network filer super cluster

Techniques for adaptive trace logging include, in one embodiment, obtaining input data on trace logging behavior and computing resources used by trace logging. Based on the obtained input data, an adaptive trace logging module automatically takes action at runtime to reduce the amount of computing resources consumed by tracing logging. For example, the action taken may include decreasing a trace logging level of an executing software program to reduce the number of trace logging messages added to a trace log. In another embodiment, the techniques include detecting a condition of an executing software program that warrants a change to a trace logging level of the executing program. The adaptive trace logging module automatically changes the trace logging level of the executing program as-needed for the detected condition. For example, the adaptive trace logging module may increase the trace logging level of an executing program upon detecting a deadlock or other abnormal condition of the executing program. By automatically increasing the trace logging level upon detecting an abnormal condition, additional trace logging messages may be written to a trace log aiding diagnosis and troubleshooting of the condition.

Techniques for adaptive trace logging

Techniques for mastering resources in a cluster of nodes are provided. A global backup lock manager (GBLM) is maintained for a cluster of nodes that implement distributed lock management. Before a server instance is taken down, for example, for maintenance purposes, such as installing a new version of the server instance code, the mastership information that the server instance stores is reflected in the mastership information maintained by the GBLM. Thus, shutting down the server instance does not involve remastering the resources mastered by the server instance. As a result, shutting down the server instance may take minimal time.

Global backup lock manager

The approaches described herein provide support for application specific policies for conventional operating systems. In an embodiment, a kernel module representing a kernel subsystem is executed within an operating system's kernel. The kernel subsystem may be configured to respond to particular requests with one or more default actions. Additionally, the kernel subsystem may define a number of sub-modules which represent application specific policies that deviate from the default actions. Each sub-module may define one or more sets of conditions which indicate when the sub-module is applicable to a request and one or more sets of corresponding actions to take when the conditions are met. When an application sends a request to the kernel subsystem, the kernel subsystem determines whether the request meets the conditions of a particular sub-module. If the particular sub-module's conditions are met, the kernel subsystem performs the corresponding actions of the particular sub-module.

Method and system for the support of application specific policies for conventional operating systems

A method for locking resources, including: receiving, by an accelerator, a first request from a first client to lock a first resource; evaluating, by a computer processor of a server, a hash function using an identifier of the first resource as an input to the hash function; identifying, by the computer processor and based on evaluating the hash function, a first hash bucket in a shared memory residing in a physical memory of the server; detecting that the first hash bucket is occupied; and sending the first request to a master lock monitor residing in a user space of the server based at least on detecting that the first hash bucket is occupied.

Lock acceleration

A method, system, and computer program product for analyzing performance of a database cluster. Disclosed are techniques for analyzing performance of components of a database cluster by transforming many discrete event measurements into a time series to identify dominant signals. The method embodiment commences by sampling the database cluster to produce a set of timestamped events, then pre-processing the timestamped events by tagging at least some of the timestamped events with a semantic tag drawn from a semantic dictionary and formatting the set of timestamped events into a time series where a time series entry comprises a time indication and a plurality of values corresponding to signal state values. Further techniques are disclosed for identifying certain signals from the time series to which is applied various statistical measurement criteria in order to isolate a set of candidate signals which are then used to identify indicative causes of database cluster behavior.

Analyzing database cluster behavior by transforming discrete time series measurements

Described herein are techniques for dynamically monitoring and managing resource usages of processes running on a node in a multi-node database system. High resource usages of processes can be proactively detected and alleviated, thereby making such a node to perform significantly better than otherwise.

Intelligent flood control management

Systems, methods, and other embodiments associated with avoiding resource blockages and hang states are described. One example computer-implemented method for a computing system includes determining that a first process is waiting for a resource and is in a blocked state. The resource that the first process is waiting for is identified. A blocking process that is holding the resource is then identified. A priority of the blocking process is compared with a priority the first process. If the priority of the blocking process is lower than the priority of the first process, the priority of the blocking process is increased. In this manner the blocking process can be scheduled for execution sooner and thus release the resource.

Global avoidance of hang states via priority inheritance in multi-node computing system

Techniques for monitoring resources of a computer system are provided. A monitoring process collects and reports utilization data for one or more resources of a computer system, such as CPU, memory, disk I/O, and network I/O. Instead of reporting just an average of the collected data over a period of time (e.g., 10 seconds), the monitoring process at least reports individually collected resource utilization values. If one or more of the utilization values exceed specified thresholds for the respective resources, then an alert may be generated. In one approach, the monitoring process is made a real-time priority process in the computer system to ensure that the memory used by the monitoring process is not swapped out of memory. Also, being a real-time priority process ensures that the monitoring process obtains a CPU in order collect resource utilization data even when the computer system is in a starvation mode.

Real-time database exception monitoring tool using instance eviction data

Growing community of inventors

Los Altos, CA, United States of America

Angelo Pruscino