Amazon Technologies, Inc.

Muhammad Ali Siddiqui

Kendra Anne Yourtee

Sanjeev Kewal Verma

Thomas Lowell Keller

Miles Childs Kaufmann

Amos Dylan Vance

Zhuo Zhang

Swaminathan Subramanian

Alexander S Borst

Gregory Suhyun Kang

Melissa Elaine Davis

Peter Varnum Commons

Arthur Carré

Gregory Richard Swensen

Ivan Eduardo Gonzalez

Charles Graham Haver Crissman

Mark Christopher Veronda

Josh M Goodspeed

Mathias G Ricken

Aaron T Olds

Joshua M Goodspeed

Colin Bodell

Rishikesan Rangarajan

Srikant Meka

Jeff M Bilger

Techniques are described for managing program execution capacity used to execute software programs, such as for a group of computing nodes in use executing one or more programs for a user. Dynamic modifications to the program execution capacity of the group may include adding or removing computing nodes, such as based on automated determinations that previously specified criteria are currently satisfied. The described techniques further include tracking information about changes that occur in one or more capacity utilization metrics of interest in response to dynamic modifications to the program execution capacity of the group, and then using such learned responses to determine a customized capacity modification policy that is configured for the actual operation of the software program(s) for the user. In some situations, the techniques are used in conjunction with a fee-based program execution service that executes multiple programs on behalf of multiple users of the service.

Learned configuration of modification policies for program execution capacity

A method is provided for estimating past data by identifying a high frequency data set for a defined time period. A pattern is calculated for the high frequency data set and then the pattern is applied to a low frequency data set in a past time period to estimate a high frequency query point.

Data estimation

Systems and methods are provided for analyzing operating metrics of monitored metric sources. Aspects of the present disclosure may present for display information associated with the monitored metric source and the analysis of its operating metrics. Analysis comprises determination of reference values and tolerance levels which represent allowable deviations from the reference values. Input data includes a measurement of an operating parameter and a time stamp. Input data may be saved to a data store for using in future analysis of other input data. When input data is determined to be outside the tolerance level, notifications may be issued to alert administrators or systems of the anomaly.

Integration based anomaly detection service

Disclosed are various embodiments for using media reported events to generate predictions for time series data. Information retrieved from a plurality of network content sources is classified into a plurality of categories. A prediction is generated for a time series. The time series is associated with a metric observed in a computing system. The generated prediction takes into account an impact of at least one of instance of the classified information.

Using media events to predict time series data

Data estimation for storing correlated patterns of high frequency data sets

The techniques described herein provide software testing that may concurrently process a user request using a live version of software and a shadow request, which is based on the user request, using a shadow version of software (e.g., trial or test version, etc.). The live version of software, unlike the shadow version, is user-facing and transmits data back to the users while the shadow request does not output to the users. An allocation module may vary allocation of the shadow requests to enable a ramp up of allocations (or possibly ramp down) of the shadow version of software. The allocation module may use allocation rules to dynamically initiate the shadow request based on various factors such as load balancing, user attributes, and/or other rules or logic. Thus, not all user requests may be issued as shadow requests.

Software testing using shadow requests

Examples of systems and methods are described for managing computing capacity by a provider of computing resources. The computing resources may include program execution capabilities, data storage or management capabilities, network bandwidth, etc. Multiple user programs can consume a single computing resource, and a single user program can consume multiple computing resources. Changes in usage and other environmental factors can require scaling of the computing resources to reduce or prevent a negative impact on performance. In some implementations, a fuzzy logic engine can be used to determine the appropriate adjustments to make to the computing resources associated with a program in order to keep a system metric within a desired operating range.

Reactive auto-scaling of capacity

Computer applications may generate event data based on a large volume of different types of record data. Described herein are systems, methods and devices for providing website recommendations using the event data. In one example, using the event data, a computing node generates the website recommendations within a designated amount of time after the generation of the record data.

Systems and methods providing recommendation data

This disclosure is directed in part to testing of different versions of software or software components (software versions) and analyzing results of use (e.g., user interaction) of the different software versions. The techniques described herein provide software testing that varies the allocation to enable a ramp up of allocations to/from another software version. The allocation module may use allocation rules to assign requests to each software version based on various factors such as load balancing, user attributes, past user assignment, and/or other rules or logic. An analysis of the different software versions may include an analysis of system performance resulting from operation of each software version. An analysis may determine attributes of each user and then allocate the user to a software version based on at least some of the determined attributes.

Software testing analysis and control

Computer applications may generate a large volume of different types of record data. In one example, the large volume of record data may represent millions of different processes occurring every second. Described herein are systems, methods and devices for generating parsed data based on the large volume of record data. The parsed data may be consumed by computing nodes within a designated amount of time from the generation of the record data.

Systems and methods providing event data

Data estimation of historical data that identifies fine grained time period data and coarse grained time period data and applies correlated slopes to time periods

Techniques are described for performing automated predictions of program execution capacity or other capacity of computing-related hardware resources that will be used to execute software programs in the future, such as for a group of computing nodes that execute one or more programs for a user. The predictions that are performed may in at least some situations be based on historical data regarding corresponding prior actual usage of execution-related capacity (e.g., for one or more prior years), and may include long-term predictions for particular future time periods that are multiple months or years into the future. In addition, the predictions of the execution-related capacity for particular future time periods may be used in various manners, including to manage execution-related capacity at or before those future time periods, such as to prepare sufficient execution-related capacity to be available at those future time periods.

Growing community of inventors

Snoqualmie, WA, United States of America

Muhammad Ali Siddiqui