Task Scheduling for Highly Concurrent Analytical and Transaction Workloads

摘要

Systems and method for a task scheduler with dynamic adjustment of concurrency levels and task granularity are disclosed for improved execution of highly concurrent analytical and transactional systems. The task scheduler can avoid both over commitment and underutilization of computing resources by monitoring and controlling the number of active worker threads. The number of active worker threads can be adapted to avoid underutilization of computing resources by giving the OS control of additional worker threads processing blocked application tasks. The task scheduler can dynamically determine a number of parallel operations for a particular task based on the number of available threads. The number of available worker threads can be determined based on the average availability of worker threads in the recent history of the application. Based on the number of available worker threads, the partitionable operation can be partitioned into a number of sub operations and executed in parallel.

主权项

1. A method comprising:
processing, by a computer system, a query in a first thread in a plurality of worker threads to generate a parsing task associated with the query; parsing, by the computer system, the query based on the parsing task to generate an execution plan comprising a plurality of ordered operations for answering the query; generating, by the computer system, a task graph based on the execution plan, the task graph comprising a plurality of task nodes corresponding to the plurality of ordered operations; identifying, by the computer system, a first task node in the plurality of task nodes corresponding to a first ordered operation in the plurality of ordered operations determined to be partitionable into sub operations for execution in parallel; generating, by the computer system, a concurrency level value that corresponds to a number of available worker threads in the plurality of worker threads; and generating, by the computer system, a plurality of partitioned tasks for execution in parallel based on the first task node and the concurrency level value.