Abstract: With the rapid development of Internet technology, human beings are moving towards the era of big data and cloud computing. As the latest generation of big data computing engine, Flink is favored by academia and industry for its advantages such as low latency and high throughput. When Flink is deployed in the cloud environment, its default task scheduling will lead to uneven load distribution due to the inability to obtain container deployment distribution information. To solve this problem, this paper proposes a Flink task scheduling load balancing algorithm in container environment to obtain the performance information of each node and the distribution information of the container on the node, give priority to the container of nodes with more free resources, and avoid the uneven load caused by the frequent selection of containers. The evaluation results show that the proposed algorithm can more evenly allocate tasks and improve resource utilization and computing speed when deployed in container environment

Key words: container, big data, Flink, task scheduling, load balancing, container environment