知识卡片
Flink面向持续流计算
Flink 以有状态流处理为核心,也能处理有界数据。对毫秒级时延或持续更新的管线,可把它列入候选;是否满足吞吐、恢复和表格式写入要求仍须用实际负载验证。
这张卡片看不懂?
这张卡片的反馈已满 10 条,暂时无法继续提交。
已有反馈 · 0 / 10 条
还没有人提交反馈。
原书出处
参考来源
- 位置:《Practical Lakehouse Architecture》第5章《Compute Engines for Lakehouse Architectures》(源文件:_epub-src/OEBPS/ch05.html)
- 结论依据:原文将 Flink 作为低时延流摄取与处理选项,官方文档确认其面向有界和无界数据流的有状态计算。
- 原始内容:Flink Spark might not be the best choice for low-latency streaming operations where you need millisecond-level latency. In such a case, Flink is another open source framework that you can use to implement streaming ingestion and processing. It offers the following features: It is a distributed, in-memory processing framework. It supports low latency, high throughput,