Abstract: Methods are provided for communicating with and regulating access to a storage system contained within a file system. Storage access keys are used to communicate to the file system changes in location of data stored in the storage system. Fence keys are used to regulate access to the storage system.

Abstract: Various techniques are described for improving the performance of a shared-nothing database system in which at least two of the nodes that are running the shared-nothing database system have shared access to a disk. Specifically, techniques are provided for changing the ownership of data in a shared-nothing database without changing the location of the data on persistent storage. Because the persistent storage location for the data is not changed during a transfer of ownership of the data, ownership can be transferred more freely and with less of a performance penalty than would otherwise be incurred by a physical relocation of the data. Various techniques are also described for providing fast run-time reassignment of ownership. Because the reassignment can be performed during run-time, the shared-nothing system does not have to be taken offline to perform the reassignment.

Abstract: Techniques are described for storing portions of files on different storage devices based on the devices' respective storage class. For example, different extents from a large file within a system, application, or database may be stored on different storage devices. The storage devices may be part of a single storage system that includes some storage devices that have different performance capabilities. The decision of where, within the storage classes, to store data may be based on the access patterns of the data. Access patterns regarding file extents are tracked using respective access bits, which are set when a memory block associated with an extent is accessed via an I/O operation. In one embodiment, an access bitmap is modified to account for cached blocks of file extents.

Abstract: Various techniques are described for improving the performance of a multiple node system by allocating, in two or more nodes of the system, partitions of a shared cache. A mapping is established between the data items managed by the system, and the various partitions of the shared cache. When a node requires a data item, the node first determines which partition of the shared cache corresponds to the required data item. If the data item does not currently reside in the corresponding partition, the data item is loaded into the corresponding partition even if the partition does not reside on the same node that requires the data item. The node then reads the data item from the corresponding partition of the shared cache.

Abstract: Locks are placed in a convert queue in a way that compensates for queue bias. Rather than always placing a remote lock in a queue at the tail, a remote lock can be placed further up in the queue, and possibly be interleaved with local locks. As a result, remote processes are granted locks more frequently and swiftly. Locks are placed in a convert queue according based on queue placement factors, which are factors accounted for when placing a lock in a queue.

Abstract: A method and mechanism is disclosed for implementing storage and compression in a computer system. Each granular portion of a file can be individually stored in either a compressed storage unit or an uncompressed storage unit. The storage units can be allocated apriori or on an as-needed basis.

Abstract: Disclosed are improved methods, systems, and mediums for handling reporting transactions in database systems. In some embodiments, database snapshots are used to carry out reporting transactions on a failover node concurrently with execution of non-reporting transactions on a primary node.

Abstract: Techniques are provided for mastering resources by assigning resources to mastery-eligible-sets. The resources that are assigned to a mastery-eligible-set may be mastered only at nodes that belong to the mastery-eligible-set. The membership of a mastery-eligible-set for a given set of related resources may be established based on which nodes have the highest affinity for the given set of related resources. Resource usage may be monitored, and the system may be periodically reconfigured, by changing membership of the mastery-eligible-sets, based on the usage information.

Abstract: Techniques are provided for mastering resources by assigning resources to mastery-eligible-sets. The resources that are assigned to a mastery-eligible-set may be mastered only at nodes that belong to the mastery-eligible-set. The membership of a mastery-eligible-set for a given set of related resources may be established based on which nodes have the highest affinity for the given set of related resources. Resource usage may be monitored, and the system may be periodically reconfigured, by changing membership of the mastery-eligible-sets, based on the usage information.

Abstract: A method and mechanism is disclosed for implementing storage and compression in a computer system. Each granular portion of a file can be individually stored in either a compressed storage unit or an uncompressed storage unit. The storage units can be allocated apriori or on an as-needed basis.

Abstract: A method and apparatus for selecting one or more instances to perform database recovery is provided. According to one aspect, one or more performance metrics are determined for each instance on each node in a cluster. Based on these performance metrics, which may be weighted to give one or more metrics more significance relative to the other metrics, an overall score is determined for each instance on each surviving node when a node in the cluster fails. One or more instances with the highest of these scores are selected. The selected instances are assigned the task of performing the database recovery process. Because instances best able to handle to database recovery process are selected, the database recovery process consumes less time, thereby making data accessible in less time after a node fails.

Abstract: Runtime load balancing of work across a clustered computing system involves servers calculating, and clients utilizing, current service performance grades of each instance in the system. A performance grade for an instance is based on performance metrics for that instance, where the computation used may vary by policy. Examples of possible policies include: (a) using estimated bandwidth as a performance grade, (b) using spare capacity as a performance grade, or (c) using response time as a performance grade. Clients distribute work requests across servers in the system as the requests arrive. Work requests can be distributed according to performance grades, and/or flags associated with the performance grades. Automatically and intelligently directing work requests to the best server instances, based on real-time service performance metrics, minimizes the need to manually relocate work within the clustered system.

Abstract: Techniques are provided for remastering resources based on node utilization. According to one such technique, resources are remastered in response to the over-utilization of the node that currently masters those resources. The utilization of each node is tracked, and when a particular node's utilization exceeds a specified threshold, selected resources that are currently mastered by that node are remastered so that nodes other than the particular node become the new masters for the selected resources. Each node's utilization is based on that node's capacity, and each node's capacity may differ. According to an aspect of one technique, each node's capacity is based on that node's processing resources and memory resources. Remastering resources in this manner tends to reduce the average amount of time taken for nodes to handle requests for the resources that they master.

Abstract: Techniques are provided for handling distributed transactions in shared-nothing database systems where one or more of the nodes have access to a shared persistent storage. Rather than coordinate the distributed transaction using a two-phase commit protocol, the coordinator of the distributed transaction uses a one-phase commit protocol with those participants that have access to the transaction status information maintained by the coordinator. The transaction status information may reside, for example, in the redo log of the coordinator. In the case that the coordinator fails, those participants can determine the state of the distributed transaction based on information stored on the shared disk. In addition, the coordinator is able to determine whether it is possible to commit the distributed transaction based on information that is stored on the shared disk by the participants, without those participants entering a formal “prepared” state.

Abstract: Various techniques are described for improving the performance of a shared-nothing database system in which at least two of the nodes that are running the shared-nothing database system have shared access to a disk. Specifically, techniques are provided for changing the ownership of data in a shared-nothing database dynamically, based on factors such as which node would be the most efficient owner relative to the performance of a particular operation. Once determined, the ownership of the data may be changed permanently to the new owner, or temporarily for the duration of the particular operation.

Abstract: Various techniques are described for improving the performance of a shared-nothing database system in which at least two of the nodes that are running the shared-nothing database system have shared access to a disk. Specifically, techniques are provided for recovering the data owned by a failed node using multiple recovery nodes operating in parallel. The data owned by a failed node is reassigned to recovery nodes that have access to the shared disk on which the data resides. The recovery logs of the failed node are read by the recovery nodes, or by a coordinator process that distributes the recovery tasks to the recovery nodes.

Abstract: Various techniques are described for improving the performance of a multiple node system by allocating, in two or more nodes of the system, partitions of a shared cache. A mapping is established between the data items managed by the system, and the various partitions of the shared cache. When a node requires a data item, the node first determines which partition of the shared cache corresponds to the required data item. If the data item does not currently reside in the corresponding partition, the data item is loaded into the corresponding partition even if the partition does not reside on the same node that requires the data item. The node then reads the data item from the corresponding partition of the shared cache.

Abstract: Methods are provided for communicating with and regulating access to a storage system contained within a file system. Storage access keys are used to communicate to the file system changes in location of data stored in the storage system. Fence keys are used to regulate access to the storage system.

Abstract: Techniques are provided for handling distributed transactions in shared-nothing database systems where one or more of the nodes have access to a shared persistent storage. Rather than coordinate the distributed transaction using a two-phase commit protocol, the coordinator of the distributed transaction uses a one-phase commit protocol with those participants that have access to the transaction status information maintained by the coordinator. The transaction status information may reside, for example, in the redo log of the coordinator. In the case that the coordinator fails, those participants can determine the state of the distributed transaction based on information stored on the shared disk. In addition, the coordinator is able to determine whether it is possible to commit the distributed transaction based on information that is stored on the shared disk by the participants, without those participants entering a formal “prepared” state.

Abstract: Various techniques are described for improving the performance of a shared-nothing database system in which at least two of the nodes that are running the shared-nothing database system have shared access to a disk. Specifically, techniques are provided for changing the ownership of data in a shared-nothing database without changing the location of the data on persistent storage. Because the persistent storage location for the data is not changed during a transfer of ownership of the data, ownership can be transferred more freely and with less of a performance penalty than would otherwise be incurred by a physical relocation of the data. Various techniques are also described for providing fast run-time reassignment of ownership. Because the reassignment can be performed during run-time, the shared-nothing system does not have to be taken offline to perform the reassignment.