Abstract: Method of managing workflow failures in a distributed computing environment. A retry value is associated with one or more workflow elements. An element or section of a workflow that does not successfully execute in the distributed computing network is identified. The workflow that does not successfully execute is retried according to a first retry value. The maximum number of times that the first element can be retried is indicated or represented by the first retry value. If one or more workflow elements fail, parent workflows can be retried according to a parent workflow retry value. Elements of the workflow that failed and successfully executed can be retried when the workflow is retried. Execution of workflow elements can also depend upon a condition concerning one or more other elements.

Abstract: Method and system for managing a distributed computing environment. The methods and systems include handling multiple heterogeneous dispatch systems, preventing deadlock in single threaded servers, optimizing distributed activities, homogeneous identification of heterogeneous resources and automatically distributing failed tasks within a distributed system.

Abstract: Method and system for gathering and propagating statistical information about resources in a distributed computing grid. Data relating to a resource in the first group of resources on the distributed computing grid is received by a gatherer. The received data is provided to other resources in the first group, and a statistical model is determined or generated for each resource in the first group based on the received data. A second group of resources on the grid is called, and the statistical information from the first group is propagated to the second group.

Abstract: Method and system for detecting indeterminate dependencies in a distributed computing grid. A determination is made whether a deadlock situation exists within a workflow of the distributed computing grid and a user of the computing grid is notified of the deadlock situation, e.g., where in the workflow deadlock occurs. A deadlock situation can be determined based on direct and indirect dependencies, such as loops and dependencies involving a first work element and a lower level second work element. A deadlock situation can also be determined based on the relationships between a job and a task, which is executable by a processor in the distributed computing grid.

Abstract: Method and system for locating objects in a computing grid. Requests for one or more objects are received from a client or other requesting entity. A persistent object locator is distributed across the grid and searches the grid for a reference to or handle on the requested object. The persistent object locator includes an internal object locator element (IOL) that resides locally, e.g., on the client, and an external object locator element (EOL) that resides on one or more other servers that are external relative to the client. The EOL is initiated if the IOL cannot locate the requested object locally. The EOL searches the grid for the requested object and if the object is available, delivers the reference to or handle on the requested object to the requesting entity. Objects can also be registered to the POL for future use. References to objects can also be deleted from the POL in the event of a communication failure or if a server or process terminates as expected.