Abstract: A computer running a database system receives one or more queries, each query comprised of parallel threads of execution working towards the common goal of completing a user request. These threads are grouped into a schedulable object called a task group. The task groups are placed within a specific multiple tier hierarchy, and database system resources allocated to the task groups according to their placement within the hierarchy. Beginning with the top tier of the hierarchy, resources remaining after allocations to each task group within a tier are passed to the next lower tier for allocation.

Abstract: A system and method to facilitate cache management and improve disk read performance for database systems with large memory and large disks. A contiguous read feature is employed to read multiple pseudo-contiguous data blocks in one large I/O from disk storage into cache memory. The contiguous read feature loads the disk area containing pseudo-contiguous data blocks by issuing a single disk read. A separate virtual space and memory page list is created for each data block, and the page lists are reunited to create one I/O. The pseudo-contiguity of two data blocks is determined by comparing the distance between them, i.e., the size of the hole between the two data blocks, with a predefined maximum distance, over which it is more effective to read the data blocks independently.

Abstract: A method, data structure, and database system for synchronizing access to a global resource by a plurality of processes executing on a computer system are disclosed. A global lock is defined for the global resource and a local instantiation of the global lock is defined for each of the plurality of processes that will access the global resource.

Abstract: Methods, computer programs, information handling systems, and state machines for performing an atomic write to a data block area are disclosed. The atomic write is an in-place write> The method includes receiving one or more data blocks to write to the data block area; and for each data block received: writing the data block to the depot slot; and writing the data block to the data block area after the data block write to the depot slot is completed.

Abstract: One or more embodiments of the invention provide for a SEG (segment) subsystem of a parallel database extension (PDE) that provides services for creating, manipulating, and destroying access to data segments in address space. The SEG subsystem works with transient segments (also referred to as scratch segments) that are created as offsets in a “well-known” file created during initialization of the PDE. The offsets are stored in objects/elements in a queue in the virtual address space of the SEG subsystem. Upon partitioning addressable memory into one or more scratch segments, obtaining the offsets for the scratch segments, and mapping the offsets into the SEG system's virtual address space, the physical memory is locked. When the scratch segments are needed, an element/object from the queue containing the appropriate size scratch segment is mapped into the requestor's address space.

Abstract: A File SeGment (FSG) subsystem of a PDE (parallel database extension) provides services for managing and allocating secondary memory such as cache, providing data commits and retrieval operations on disk, and providing routines to perform disk I/O (Input/Output) and other memory operations for a file system. The FSG subsystem provides for the determination of the amount of cache that may be utilized, locking that portion of cache, using a buddy memory management system to manage the cache, using one or more hash aging queues to determine when to free up cache, permitting preloads of data into the cache, and asynchronously accepting an I/O completion action from a user prior to completion of an operation and performing such an I/O completion action.

Abstract: A processor method and apparatus for parallel computer systems for synchronizing the flushing of database segments by a node through shared disk tokens. The system provides for flushing database segments at each of the processor nodes through use of a persistent storage semaphore stored as a plurality of disk tokens and a set of control flags in a synchronization segment on a shared data storage device.