Abstract:
A technique for resending messages can include determining if a particular message with a first unique identifier has been received by a recipient task among a number of potential recipient tasks in a distributed streaming system, notifying the number of potential recipient tasks to ignore the particular message with the first unique identifier, and resending the particular message with a second unique identifier.
Abstract:
A technique of recovering a failure in a data processing system comprises restoring a checkpointed state in a last window, and resending all the input messages received at the second node during the failed window boundary.
Abstract:
Described herein are techniques enabling a query engine to process a query comprising a user defined function. The user defined function can include a request for analytics to be performed by an external analytic engine. The query engine can retrieve data from a database in accordance with the query and send the data and instructions to perform the analytics to the external analytic engine. The query engine can then receive analytic results from the external analytic engine.
Abstract:
Multiple tuples are combined (202) by a processing node into a message that has a key portion including a value of a key, and a payload including a relation containing content of the combined plurality of tuple. The message is sent (204) to a receiving node.
Abstract:
A technique for resending messages can include determining if a particular message with a first unique identifier has been received by a recipient task among a number of potential recipient tasks in a distributed streaming system, notifying the number of potential recipient tasks to ignore the particular message with the first unique identifier, and resending the particular message with a second unique identifier.
Abstract:
Query integration across databases and file systems is disclosed. An example method may include streaming data managed by a first database file system for a query. The method may also include streaming data managed by a second database file system for the query. The method may also include joining the streaming data managed by the first database file system with the streaming data managed by the second database file system.
Abstract:
Processing streaming data with open executors includes receiving input data at a computation dataflow station where the computation dataflow station contains a computation file and an open executor that accepts code plug-ins, converting contents of the computation file into a program string with the code plug-ins from a system library, and launching the program string together with the input data to calculate an output with a graphics processing unit.
Abstract:
According to an example, a dataflow directed acyclic graph including nodes and edges connecting the nodes. Parameters for the nodes are calculated, and the parameters may include batch size and a degree of parallelism.
Abstract:
Described herein are techniques for handling a query that uses first data from a first database and second data from a second database. The first database is accessible via a first execution engine, and the second database is accessible via a second execution engine. A sub-query of the query can be sent to the second execution engine for execution on the second database. Streaming results of the sub-query can be received by the first execution engine.