Abstract: An apparatus and method for a query governor to govern query execution in a parallel computer system with one or more networks. The query governor uses a network monitor to gather network status information to determine how to govern query execution. The query governor may also use historical information of previous query execution stored in a query file to determine if the query should be allowed to execute. The query governor may also use node and network restrictions to determine what networks the query can use.

Abstract: A database spread over multiple nodes allows each node to store a journal recording changes made to the database and also allows a journaling component to manage the memory space available for journaling. Two threshold size values may be specified for the journal. The first threshold value specifies a journal size at which to being pruning the journal on a given node. A journal pruning algorithm may be used to identify journal entries that may be removed. For example, once a given transaction completes (i.e., commits) the journal entries related to that transaction may be pruned from the journal. The second threshold value specifies the maximum size of the journal. After reaching this size, journal entries may be written to disk instead of the in-memory journal.

Abstract: A database spread over multiple nodes allows each node to store a journal recording changes made to the database and also allows a journaling component to manage the memory space available for journaling. Two threshold size values may be specified for the journal. The first threshold value specifies a journal size at which to being pruning the journal on a given node. A journal pruning algorithm may be used to identify journal entries that may be removed. For example, once a given transaction completes (i.e., commits) the journal entries related to that transaction may be pruned from the journal. The second threshold value specifies the maximum size of the journal. After reaching this size, journal entries may be written to disk instead of the in-memory journal.

Abstract: Embodiments of the present invention generally provide for network virtualization in a multi-network system. The virtual network may abstract multiple, underlying physical networks connecting compute nodes on a multi-node system. A network driver and control system may route traffic on the underlying physical networks according to system and/or application requirements, and a virtual-to-physical network mapping may be controlled by the driver and control system.

Abstract: Systems and articles of manufacture for managing annotations made for a variety of different type data objects manipulated (e.g., created, edited, and viewed) by a variety of different type applications are provided. Some embodiments allow users collaborating on a project to create, view, and edit annotations from within the applications used to manipulate the annotated data objects, which may facilitate and encourage the capturing and sharing of tacit knowledge through annotations. Further, annotations may be stored separate from the application data they describe, decoupling the tacit knowledge captured in the annotations from the applications used to manipulate the annotated data.

Abstract: Techniques are disclosed for reducing the number of two-phase commits required to perform atomic transactions in a multi-node system. By monitoring atomic operations that require two-phase commit protocols, a system may identify when it would be beneficial to consolidate data onto single (or at least fewer) compute node(s). In doing so, a less compute intensive commit protocol, such as a one-phase commit protocol may be used. Moreover, in situations where data consolidation is not available, the data may be migrated to compute nodes having a closer proximity. Thus, in such cases, network response times for two-phase commits required for an atomic operation may be reduced, thereby increasing the performance of a multi-node system.

Abstract: A method, apparatus, and program product checkpoint an application in a parallel computing system of the type that includes a plurality of hybrid nodes. Each hybrid node includes a host element and a plurality of accelerator elements. Each host element may include at least one multithreaded processor, and each accelerator element may include at least one multi-element processor. In a first hybrid node from among the plurality of hybrid nodes, checkpointing the application includes executing at least a portion of the application in the host element, configuring and executing at least one computation kernel in at least one accelerator element, and, in response to receiving a command to checkpoint the application, checkpointing the host element separately from the at least one accelerator element upon which the at least one computation kernel is executing.

Abstract: In one aspect of the invention, a method is provided. The method may include: (1) storing a snapshot of a system state of a node; (2) executing a job on the node; and (3) restoring the node to the system state using the stored snapshot of the system state.

Abstract: Embodiments of the invention may be used to increase query processing parallelism of an in-memory database stored on a parallel computing system. A group of compute nodes each store a portion of data as part of the in-memory database. Further, a pool of compute nodes may be reserved to create copies of data from the compute nodes of the in-memory database as part of query processing. When a query is received for execution, the query may be evaluated to determine whether portions of in-memory should be duplicated to allow multiple elements of the query (e.g., multiple query predicates) to be evaluated in parallel.

Abstract: Embodiments of the invention provide techniques for scheduling jobs on a multi-node computing system based on the predicted environmental impact of executing the jobs. In one embodiment, a plurality of job plans may be generated for processing a requested job on the multi-node computing system. The environmental impacts resulting from executing each job plan may be estimated by matching the job plans to stored data based on standardized executions of job plans. Further, environmental impacts may be estimated by matching the job plans to stored data based on actual environmental measurements obtained during prior executions of the job plan on the multi-node computer system. The job may be executed using a job plan selected based on predicted environmental impacts and time performance.

Abstract: A method, apparatus, and program product optimize power consumption in a parallel computing system that includes a plurality of computing nodes by selectively throttling performance of selected nodes to effectively slow down the completion of quicker executing parts of a workload of the computing system when those parts are dependent upon or otherwise associated with the completion of other, slower executing parts of the same workload. Parts of the workload are executed on the computing nodes, including concurrently executing a first part on a first computing node and a second part on a second computing node. The first node is selectively throttled during execution of the first part to decrease power consumption of the first node and conform a completion time of for the first node in completing the first part of the workload with a completion time for the second node in completing the second part.

Abstract: A method, apparatus, and program product manage scheduling of a plurality of jobs in a parallel computing system of the type that includes a plurality of computing nodes and is disposed in a data center. The plurality of jobs are scheduled for execution on a group of computing nodes from the plurality of computing nodes based on the physical locations of the plurality of computing nodes in the data center. The group of computing nodes is further selected so as to distribute at least one of a heat load and an energy load within the data center. The plurality of jobs may be additionally scheduled based upon an estimated processing requirement for each job of the plurality of jobs.

Abstract: A parallel computer system has multiple nodes that have independent clocks, where the different nodes may include different database portions that are referenced by a query. A timestamp parameter in a query is synchronized across the different nodes that are referenced by the query to assure the timestamps in the different nodes are consistent with each other notwithstanding the independent clocks used in each node. As a result, a database may be scaled to a parallel computer system with multiple nodes in a way that assures the timestamps for different nodes referenced during a query have identical values.

Abstract: Embodiments of the invention provide techniques for presenting energy consumption information in an IDE tool. In one embodiment, the IDE tool may be configured to determine energy requirements associated with specific elements of the source code, and to present graphical indications of energy requirements along with those code elements. Such energy requirements may be determined by matching code elements to a stored data structure describing energy requirements for executing various code elements. The stored data may be based on predefined standards, and/or may be based on historical measurements of energy consumption during prior instances of executing code elements on a target computer system. Additionally, developers may specify priority for compiling portions of source code, according to desired energy requirements.

Abstract: Embodiments of the invention provide techniques for presenting energy consumption information in an IDE tool. In one embodiment, the IDE tool may be configured to determine energy requirements associated with specific elements of the source code, and to present graphical indications of energy requirements along with those code elements. Such energy requirements may be determined by matching code elements to a stored data structure describing energy requirements for executing various code elements. The stored data may be based on predefined standards, and/or may be based on historical measurements of energy consumption during prior instances of executing code elements on a target computer system. Additionally, developers may specify priority for compiling portions of source code, according to desired energy requirements.

Abstract: Embodiments of the invention provide techniques that improve resource management on a massively parallel computing system having a plurality of hybrid compute nodes. For example, a job scheduler may be provided which determines a library to link to an application based on system and user requirements. In one embodiment, the libraries may provide optimizations for job execution time, and also provide optimizations directed towards a specific processor architecture. Once the library is determined, the job scheduler may configure the environment of the application so that the application links with the optimized library during run-time. Doing so may improve overall system performance of the massively parallel computing system.

Abstract: A method for optimizing efficiency and power consumption in a hybrid computer system is disclosed. The hybrid computer system may comprise one or more front-end nodes connected to a multi-node computer system. Portions of an application may be offloaded from the front-end nodes to the multi-node computer system. By building historical profiles of the applications running on the multi-node computer system, the system can analyze the trade offs between power consumption and performance. For example, if running the application on the multi-node computer system cuts the run time by 5% but increases power consumption by 20% it may be more advantageous to simply run the entire application on the front-end.

Abstract: An apparatus and method for overcoming a torus network failure in a parallel computer system. A mesh routing mechanism in the service node of the computer system configures the nodes from a torus to a mesh network when a failure occurs in the torus network. The mesh routing mechanism takes advantage of cutoff registers in each node to route node to node data transfers around the faulty node or network connection.

Abstract: Techniques are disclosed for reducing the number of two-phase commits required to perform atomic transactions in a multi-node system. By monitoring atomic operations that require two-phase commit protocols, a system may identify when it would be beneficial to consolidate data onto single (or at least fewer) compute node(s). In doing so, a less compute intensive commit protocol, such as a one-phase commit protocol may be used. Moreover, in situations where data consolidation is not available, the data may be migrated to compute nodes having a closer proximity. Thus, in such cases, network response times for two-phase commits required for an atomic operation may be reduced, thereby increasing the performance of a multi-node system.

Abstract: Efficient application checkpointing uses checkpointing characteristics of a job to determine how to schedule jobs for execution on a multi-node computer system. A checkpoint profile in the job description includes information on the expected frequency and duration of a check point cycle for the application. The checkpoint profile may be based on a user/administrator input as well as historical information. The job scheduler will attempt to group applications (jobs) that have the same checkpoint profile, on the same nodes or group of nodes. Additionally, the job scheduler may control when new jobs start based on when the next checkpoint cycle(s) are expected. The checkpoint monitor will monitor the checkpoint cycles, updating the checkpoint profiles of running jobs. The checkpoint monitor will also keep track of an overall system checkpoint profile to determine the available checkpointing capacity before scheduling jobs on the cluster.