Abstract: A method and apparatus for write-back caching in a data storage and processing system has been described. The method comprises the steps of receiving a write request including write data from a compute node in a first I/O node, forwarding the write data from the first I/O node to a second I/O node, and sending an acknowledgment message to the compute node from the second I/O node after the write data is received by the second I/O node. After the data is written into non-volatile storage of the first I/O node, a purge request or command is sent to the second I/O node to purge the write data from the volatile memory of the second I/O node. In one embodiment, the purge request is not sent until the first I/O node receives a second write request, in which case, the purge request is sent in the same interrupt as the write data for the second write request.

Abstract: A method and apparatus for communicating data in a highly distributed parallel processing computer architecture is described. The method comprises the steps of generating a globally unique ID in the I/O node for a data extent physically stored in the plurality of storage devices, binding the globally unique ID to the data extent, and exporting the globally unique ID to the compute nodes via the interconnect fabric. In one embodiment, the globally unique ID is generated from a globally unique I/O node identifier and a locally unique data extent identifier. A local entry point is generated in the compute node for the data associated with the globally unique ID, thereby presenting the globally unique ID as a device point in the compute node.

Abstract: A highly-scalable parallel processing computer system architecture is described. The parallel processing system comprises a plurality of compute nodes for executing applications, a plurality of I/O nodes, each communicatively coupled to a plurality of storage resources, and an interconnect fabric providing communication between any of the compute nodes and any of the I/O nodes. The interconnect fabric comprises a network for connecting the compute nodes and the I/O nodes, the network comprising a plurality of switch nodes arranged into more than g(logbN) switch node stages, wherein b is a total number of switch node input/output ports, and g(x) indicates a ceiling function providing the smallest integer not less than the argument x, the switch node stages thereby providing a plurality of paths between any network input port and network output port.

Abstract: The present invention describes a parallel processing system. The system comprises a plurality of compute nodes for executing applications via a storage application interface having system input/output calls, a plurality of I/O nodes, and a file system implemented in the compute node, for storing information mapping API system input/output calls for the data object with the globally unique identification for the data object. Each I/O node manages a communicatively coupled plurality of storage resources and each has a means for generating a globally unique identification for a data object stored on the storage resource and transmits the globally unique identification and the data object to the compute node via at least one interconnecting fabric providing communication between any of the compute node and any of the I/O nodes.