Microsoft Technology Licensing, LLC

Derek Kiwamu Tsuchida

Baskaran Dharmarajan

Jeffrey Earl Steinbok

Kevin Thomas Brown

Jeremy Haubold

Harry Phillip Ferguson, Iv

Shabbir A Shahpurwala

Archana Manoj Gadkari

Joel Scaria

Shardul Vikram

Parker Shelton

Sergio Ligregni

Amar Shroff

Kyle Richey

Techniques are disclosed for allocating computing resources. Historical verification test data for one or more workloads is accessed. The accessed data is updated by removing data for workloads that are mathematical outliers. Based on the updated data, for one or more workload groups, resource allocations for a range of test times are determined. Based on the allocations, costs for the one or more workload groups are predicted. One or more test configurations are updated based on the predicting.

Optimization of workloads based on constraints

Systems and methods for creating training data for a software application that executes automated tasks such as operating system shell. One example provides a system that comprises at least one electronic processor configured to access a file repository that contains programming files. The electronic processor is also configured to extract data from documentation associated with at least one of the programming files to generate extracted data, extract attributes associated with at least one of the programming files to generate extracted attributes, convert the extracted data and the extracted attributes to training data, and store the training data in a training data repository. The electronic processor is also configured to update the documentation and attributes in the programming files (for example, by providing a user interface to allow a user to do so or in another fashion).

Natural language interface to interactive, operating-system shell and techniques for creating training data for the same

Systems and methods for providing user's access to a particular version of an electronic resource (e.g., a website, web resource or the like) where versions of such electronic resources are stored across a set of servers are disclosed. In one embodiment, user's requests may be received&#x2014;either requesting a particular version or as an unversioned request. A version control module (for example, a load balancer) may receive these requests and assign the user's request to a first server according to different metrics, e.g., regarding version control rules and/or effective load balancing considerations. If the initial server assigned is not able to handle the user's request, the user's request may be proxied to another server, according to different metrics. If there is no server that may handle the user's request (after a certain number of proxied requests), the request may be returned to the user as not handled.

Dynamically affinitizing users to a version of a website

Growing community of inventors

Kirkland, WA, United States of America

Derek Kiwamu Tsuchida