Task scheduling method and device

Abstract

The embodiment of the present application provides a task scheduling method and a device, which belong to the field of data processing. The method comprises the following steps of: acquiring the actual number of messages processed by each of at least two cores of a multi-core processor in a preset period of time; according to the actual number of messages processed by each core in the preset timeperiod, scheduling at least one task block in a first core of the at least two cores to a second core, wherein a first actual number of messages processed by the first core in the preset period of time is greater than a second actual number of messages processed by the second core in the preset period of time. In this scheme, the scheduling of task blocks among multiple cores is realized by acquiring the actual number of processed packets of each core, so that the effective running workload of each core task can be more accurately counted, the tasks can be distributed among multiple cores moreeffectively, and the optimal performance of the whole system can be achieved.

Description

technical field [0001] The present application relates to the field of data processing, and in particular, to a task scheduling method and device. Background technique [0002] In Network Function Virtualization (NFV for short) systems, in order to achieve the ultimate performance of data processing, most of the following technologies are used: polling processing is used instead of interrupt notification to eliminate interrupt overhead; each core of the multi-core processor Bind only one polling processing thread to eliminate task scheduling overhead among multiple cores. However, some processing steps in the data processing flow may not necessarily be assigned to fixed cores, such as Qos dequeue operation, there may be thousands of queues that need to be dequeued, and these dequeue processes need to be assigned to different cores. The usual practice is to divide these different processing steps in the message processing flow into basic task blocks, which are collectively c...

AI Technical Summary

Problems solved by technology

[0003] However, the load of each core calculated by the existing method is the same, so the jobs in each core cannot be scheduled, and in practical applications, the load of each core is unbalanced, so, It is impossible to achieve load balancing of each core by implementing job scheduling through existing methods.

Images