Abstract: A method for executing database transactions is disclosed. A plurality of interconnected nodes are each defined in terms of processor and storage resources of a parallel computing system. A first set of virtual processors is mapped across a first subset of the nodes to create a first map with at least one virtual processor being mapped to each node in the first subset. A second set of virtual processors is mapped across a second subset of the nodes to create a second map with at least one virtual processor being mapped to each node in the second subset. The first map is stored as a first configuration and the second map is stored as a second configuration. At least one transaction is executed using the first set of virtual processors and simultaneously at least one transaction is executed using the second set of virtual processors.

Abstract: A method, apparatus, and article of manufacture for coordinating a plurality of sub-tasks performed by a group of nodes of a parallel processor computer system. An application subdivides a function into the plurality of sub-tasks and assigns the sub-tasks to a group of participating nodes. A message is transmitted by the application to the group of participating nodes, wherein the message spawns one or more threads on each participating node and the threads are encapsulated by a virtual processor to perform a desired one or more of the sub-tasks. A channel object is created in all participating nodes and a coordinator node in response to the transmitted message, wherein the channel object in the coordinator node includes a data element identifying the group of participating nodes performing the sub-tasks.

Abstract: A method, apparatus, and article of manufacture for sorting data on a parallel processing computer system, wherein the system is comprised of one or more nodes, each of the nodes executes one or more threads to generate a subset of the data, and the subset of data comprises one or more records. A local data structure is generated on each of the nodes, wherein the local data structure stores ordered entries associated with the records generated by each of the threads. A global data structure is generated on a coordinator node, wherein the global data structure stores ordered entries associated with the records stored in each of the local data structures generated by each of the nodes. An answer set is generated using sorted records from the global data structure.