Microsoft Technology Licensing, LLC

Jun Su

Mingqiang Xu

Tianchi Ma

Salim Alam

Colin Watson

Gregory Marshall Burgess

Yiding Zhou

Rich Ciapala

Zhen Wei

Arun Mathew Abraham

Tao Huang

Jing Fan

Peter Gerard Provost

Qiufang Shi

Jean-Marc Prieur

Lanlan Cong

Yun Jin

Haiying Xu

An embodiment includes selecting a job to upload to a remote system, uploading a portion of data for the job to the remote system, and receiving notice that the portion of data for the job has been staged at the remote system. The embodiment also includes, based on the portion of data being staged, instructing the remote system to begin performing operations for the job on the portion of the data, where after the remote system begins to perform the operations for the job on the portion of data prior to all data for the job being uploaded. From an alternate perspective an embodiment includes receiving a description of a job to be performed on data, receiving a portion of the data, receiving an instruction to begin performing computations on the portion of data, and beginning to perform identified computations on the portion of data before receiving all the data.

Smart data staging based on scheduling policy

An application combines on-site and cloud-based calculations. The application selects a cloud computing broker from a list sorted by load. The broker receives computation requests from the application, sends corresponding load-balanced requests to local application service hosts and cloud proxies, receives responses, and sends the responses to the application without identifying the response source as local versus non-local. Each proxy has a logical-physical mapping of cloud-resident worker nodes. A scheduler receives session requests from applications, allocates local and cloud computing resources to a session, and tracks resource usage by session. An application previously run only on-premises can be executed partially on cloud worker nodes without modification. The proxy re-routes unfulfilled requests from a failed worker, undetected by the broker. If all workers assigned to that proxy fail, then the broker reassigns unfulfilled requests to another proxy, undetected by the application.

Multilevel multipath widely distributed computational node scenarios

The present invention extends to methods, systems, and computer program products for identifying impacted tests from statically collected data. In general, static dependency data, possibly augmented with some dynamic data, is used to find an appropriate set of impacted tests for code changes. In some embodiments, static dependency analysis is used to identify tests impacted by a code change. Heuristics can be used to assist with identifying an appropriate set of impacted tests to run for a code change. Dynamic data can be used to augment static dependency data to identify more optimal sets of impacted tests to run for a code change.

Identifying impacted tests from statically collected data

The subject disclosure is directed towards automatically managing job request execution for a computer cluster using cloud computing resource allocations. When client computers to the computer cluster submit job requests to a head node, a set of job requests is selected based on a policy. The head node converts the set of job requests into a set of resource allocation requests based on job specification data. After communicating the resource allocation requests to a cloud computing provider, the head node is granted access and control over one or more worker nodes as a response. The worker nodes proceed to execute the set of job requests and update the head node with status information once the execution finishes.

Enhanced computer cluster operation using resource allocation requests

Moving data to a remote system and performing computations at the remote system. A method includes receiving from one or more users a plurality jobs. The one or more jobs specify both data to be operated on and the operations to be performed on the data at the remote system. At a data service, one of the jobs to upload to the remote system in priority over other jobs in the plurality of jobs is selected. The method further includes beginning to upload the prioritized job including staging data from the job. At the remote system, once a portion of the data has been received, a service in the remote system begins to perform the specified operation on the data for the prioritized job. Beginning to perform the specified operation is performed prior to all of the data from the prioritized job being uploaded to the remote system.

External processing of a request directed to an external software program from an invocation source to a cluster of computing devices. A wrapper tool is provided for processing the request from the invocation source. The request includes operations to be performed by the external software program on the host computer. The wrapper tool intercepted the request and extracts information included in the intercepted request. The extracted information is transmitted to a cluster scheduler. The cluster scheduler identifies a set of computing devices in the cluster of computing devices for processing the request. The identified set of computing devices processes the intercepted request as a function of the intercepted request, the information and the invocation source. A result is provided in response to the processed request from the identified set of computing devices to the invocation source.

Transparently externalizing plug-in computation to cluster

Growing community of inventors

Shanghai, China

Jun Su