Microsoft Technology Licensing, LLC

Emad A Omara

John Joseph Duffy

Massimo Mascaro

Igor Ostrovsky

Stephen Harris Toub

A method of executing an algorithm in a parallel manner using a plurality of concurrent threads includes generating a lock-free barrier that includes a variable that stores both a total participants count and a current participants count. The total participants count indicates a total number of threads in the plurality of concurrent threads that are participating in a current phase of the algorithm, and the current participants count indicates a total number of threads in the plurality of concurrent threads that have completed the current phase. The barrier blocks the threads that have completed the current phase. The total participants count is dynamically updated during execution of the current phase of the algorithm. The generating, blocking, and dynamically updating are performed by at least one processor.

Lock-free barrier with dynamic updating of participant count using a lock-free technique

A dataflow of a distributed application is visualized in a locally simulated execution environment. A scheduler receives a job graph which includes a graph of computational vertices that are designed to be executed on multiple distributed computer systems. The scheduler queries a graph manager to determine which computational vertices of the job graph are ready for execution in a local execution environment. The scheduler queries a cluster manager to determine the organizational topology of the distributed computer systems to simulate the determined topology in the local execution environment. The scheduler queries a data manager to determine data storage locations for each of the computational vertices indicated as being ready for execution in the local execution environment. The scheduler also indicates an instance of each computational vertex to be spawned and executed in the local execution environment based on the organizational topology and indicated data storage locations.

Simulation environment for distributed programs

An enumerable concurrent data structure referred to as a concurrent bag is provided. The concurrent bag is accessible by concurrent threads and includes a set of local lists configured as a linked list and a dictionary. The dictionary includes an entry for each local list that identifies the thread that created the local list and the location of the local list. Each local list includes a set of data elements configured as a linked list. A global lock on the concurrent bag and local locks on each local list allow operations that involve enumeration to be performed on the concurrent bag.

Enumeration of a concurrent data structure

Embodiments described herein are directed to dynamically controlling the number of spins for a selected processing thread among a plurality of processing threads. A computer system tracks both the number of waiting processing threads and each thread's turn, wherein a selected thread's turn comprises the total number of waiting processing threads after the selected thread's arrival at the processor. Next, the computer system determines, based on the selected thread's turn, the number of spins that are to occur before the selected thread checks for an available thread lock. The computer system also, based on the selected thread's turn, changes the number of spins, such that the number of spins for the selected thread is a function of the number of waiting processing threads and processors in the computer system.

Scalable thread locking with customizable spinning

Lock-free barrier with dynamic updating of participant count

A scheduler receives a job graph which includes a graph of computational vertices that are designed to be executed on multiple distributed computer systems. The scheduler queries a graph manager to determine which computational vertices of the job graph are ready for execution in a local execution environment. The scheduler queries a cluster manager to determine the organizational topology of the distributed computer systems to simulate the determined topology in the local execution environment. The scheduler queries a data manager to determine data storage locations for each of the computational vertices indicated as being ready for execution in the local execution environment. The scheduler also indicates to a vertex spawner that an instance of each computational vertex is to be spawned in the local execution environment based on the organizational topology and indicated data storage locations, and indicates to the local execution environment that the spawned vertices are to be executed.

Embodiments described herein are directed to dynamically controlling the number of spins for a selected processing thread among a plurality of processing threads. A computer system tracks both the number of waiting processing threads and each thread's turn, wherein a selected thread's turn comprises the total number of waiting processing threads after the selected thread's arrival at the processor. Next, the computer system determines, based the selected thread's turn, the number of spins that are to occur before the selected thread checks for an available thread lock. The computer system also, based on the selected thread's turn, changes the number of spins, such that the number of spins for the selected thread is a function of the number of waiting processing threads and processors in the computer system.

A method of storing per-thread, per-instance data, includes identifying a unique index value corresponding to a first instance, identifying type parameters based on the identified index value, and instantiating a generic holder object based on the identified type parameters. The generic holder object includes a thread local field configured to store per-thread data that is local to the first instance.

Growing community of inventors

Redmond, WA, United States of America

Emad A Omara