Microsoft Technology Licensing, LLC

Other

Microsoft Corporation

Eric P Traut

Rene Antonio Vega

Matthew Douglas Hendel

Forrest Curtis Foltz

Suyash Sinha

Adrian J Oney

Bryan M Willman

Aaron Sebastian Giles

Jeffrey Kinsey

Adam D Stritzel

Jacob Kappeler Oshins

John Michael Sheehan

Nathan T Lewis

Parag Chakraborty

John Te-Jui Sheu

Idan Avraham

Mike Neil

Osama M Salem

Dustin Lyle Green

Dan Teodosiu

Ravisankar V Pudipeddi

Pankaj Garg

Andrew John Thornton

Andrew Ernest Nicholas

Xiongjian Fu

Tahsin Erdogan

Adriaan W Canter

Christopher D Russell

Peter L Johnston

David S Bailey

Joy Ganguly

Marcos De Moura Campos

John R Douceur

Scott Field

Richard Bruce Ward

Ashish Kapoor

Jonathan Ryan Howell

Jacob Rubin Lorch

David N Cutler

John Andrew Starks

Shuvabrata Ganguly

Jose Bernabeu-Auban

Brandon S Baker

Mark Robbin Brown

Kevin Michael Broas

Senthil Rajaram

Dattatraya B Rajpure

Carolyn L Napier

Mathew John

Rahul Newaskar

Darryl E Havens

René A Vega

Todd William Harris

Andrey Shedel

Prasad Thiruveedu

Ross Ian Story

Victor Shnayder

Scott A Field

Babak Seyed Aghazadeh

Mohamed Bouchet

Hari Krishna Pammi

Yogesh K Sharma

Carolyn N Lesser

Renee Antonio Vega

Benjamin Martz

Xuan Zhao

Vega Rene A A

Nathan Lewis

Matthew D Hendel

A method of training a reinforcement machine learning computer system. The method comprises providing a machine-learning computer programming language including a pre-defined plurality of reinforcement machine learning criterion statements, and receiving a training specification authored in the machine-learning computer programming language. The training specification defines a plurality of training sub-goals with a corresponding plurality of the reinforcement machine learning criterion statements supported by the machine-learning computer programming language. The method further comprises computer translating the plurality of training sub-goals from the training specification into a shaped reward function configured to score a reinforcement machine learning model configuration with regard to the plurality of training sub-goals. The method further comprises running a training experiment with the reinforcement machine learning model configuration, scoring the reinforcement machine learning model in the training experiment with the shaped reward function, and adjusting the reinforcement machine learning model configuration based on the shaped reward function.

Training reinforcement machine learning systems

A method of training a machine learning system. The method comprises collecting a first simulation dataset derived from a computer simulating a hypothetical scenario with a first simulation configuration having a first degree of fidelity. The method further comprises collecting a second simulation dataset derived from a computer simulating the hypothetical scenario with a second simulation configuration having a second degree of fidelity different than the first degree of fidelity. The method further comprises building a multi-fidelity training dataset including training data from both the first simulation dataset and the second simulation dataset according to an interleaving protocol.

Multi-fidelity simulated data for machine learning

Efficient power management of a system with virtual machines is disclosed. In particular, such efficient power management may enable coordination of system-wide power changes with virtual machines. Additionally, such efficient power management may enable coherent power changes in a system with a virtual machine monitor. Furthermore, such efficient power management may enable dynamic control and communication of power state changes.

Efficient power management of a system with virtual machines

A virtual machine monitor (VMM) is configured to enforce deterministic execution of virtual machines in a multiprocessor machine. The VMM is configured to ensure that any communication by physical processors via shared memory is deterministic. When such VMMs are implemented in a distributed environment of multiprocessor machines coupled via a logical communication link, non-deterministic server applications running on virtual machines using the VMM may be replicated.

Enforcing deterministic execution of threads of guest operating systems running in a virtual machine hosted on a multiprocessor machine

Techniques for adjusting memory in virtual machines are disclosed. According to aspects, memory status is obtained for a guest operating system. Based on the obtained memory status, an amount of guest physical addresses is reported to a memory manager of the guest operating system. Moreover, the amount memory assigned to the guest operating system may be adjusted during the runtime operation of the guest operating system.

Dynamic virtual machine memory management

In an exemplary embodiment, a virtual disk file can be assigned an identifier and a virtual disk files that is dependent on the virtual disk file can include a copy of the identifier. In the instance that the virtual disk file is opened and data is modified that causes the contents of a virtual disk extent to change the identifier can be changed. If the virtual disk file and the dependent virtual disk file are used to instantiate a virtual disk the difference between identifiers can be detected, which is indicative of the fact that the virtual disk may be corrupted. Other techniques are described in the detailed description, claims, and figures that form a part of this document.

Detecting and preventing virtual disk storage linkage faults

Systems, methods, and computer program products are described that provide a user interface to a user of a first operating system to enable the user to interact with a second operating system executing in a virtual machine. If the first and second operating systems are in communication in accordance with an operating system remote terminal (OSRT) protocol, such as Remote Desktop Protocol, a first user interface is provided as the user interface. If the first and second operating systems are not in communication in accordance with the OSRT protocol, a second user interface is provided as the user interface.

User interface for interaction with virtual machine

An installation package associated with an application is obtained. Installation of the application in a first type of system is analyzed based at least in part on the installation package. A first set of changes made to the first type of system resulting from installation of the application in the first type of system is identified. Based on the first set of changes, a second set of changes made to a second type of system resulting from installation of the application in the second type of system is identified, and a recipe identifying the second set of changes is generated. A new installation package associated with the application can also be generated. This new installation package is configured for installation of the application on the second type of system, and is based at least in part on both the recipe and the installation package.

Recipe-based application conversion

Various mechanisms are disclosed herein for the saving and restoring of virtual machine environment state. For example, virtual machine state can be either be saved or (multiple) snapshots can be taken of the virtual machine state. In the latter case, virtual processors can be allowed to run while the memory of the virtual machine state is being saved. In either case, virtual devices associated with the virtual machine environment can be quiesced such that these devices can prepare themselves to be saved. Once such virtual devices and memory are saved, they can also be restored. For example, restoration of memory can occur while virtual processors are running at the same time. And, moreover, restoration can occur in batches of pages, thus optimizing the response time for restoring saved data.

Concurrent virtual machine snapshots and restore

The present invention is directed to making a guest operating system aware of the topology of the subset of host resources currently assigned to it. At virtual machine boot time a Static Resource Affinity Table (SRAT) will be used by the virtualizer to group guest physical memory and guest virtual processors into virtual nodes. Thereafter, in one embodiment, the host physical memory behind a virtual node can be changed by the virtualizer as necessary, and the virtualizer will provide physical processors appropriate for the virtual processors in that node.

Systems and methods for exposing processor topology for virtual machines

A request is received from a software component of a software product to access a value of a public setting of the software component. A check is made as to whether the request conforms to an appropriate format for the public setting. If the request conforms to the appropriate format, then the requested access is performed. However, if the request does not conform to the appropriate format, then the requested access is denied. Multiple values for the public setting can be received and maintained concurrently.

Controlling access to software component state

Systems and methods are provided, whereby partitions may become enlightened and discover the presence of a hypervisor. Several techniques of hypervisor discovery are discussed, such as detecting the presence of virtual processor registers (e.g. model specific registers or special-purpose registers) or the presence of virtual hardware devices. Upon discovery, information (code and/or data) may be injected in a partition by the hypervisor, whereby such injection allows the partition to call the hypervisor. Moreover, the hypervisor may present a versioning mechanism that allows the partition to match up the version of the hypervisor to its virtual devices. Next, once code and/or data is injected, calling conventions are established that allow the partition and the hypervisor to communicate, so that the hypervisor may perform some operations on behalf of the partition. Four exemplary calling conventions are considered: restartable instructions, a looping mechanism, shared memory transport, and synchronous or asynchronous processed packets. Last, cancellation mechanisms are considered, whereby partition requests may be cancelled.

Systems and methods for hypervisor discovery and utilization

Growing community of inventors

Bellevue, WA, United States of America

Eric P Traut