论文标题
轨道枪：关键系统的流媒体窗口
Railgun: streaming windows for mission critical systems
论文作者
论文摘要
一些关键任务系统（例如欺诈检测）需要在需要高吞吐量和低潜伏期的应用程序上进行准确的实时指标。由于这些应用程序需要“永远”运行，因此应对大型和尖峰的数据负载，因此它们需要在分布式设置中运行。毫不奇怪，我们不知道提供所有这些属性的任何分布式流媒体系统。取而代之的是，现有系统进行大量简化，例如将滑动窗口作为固定的部分重叠窗口，危害度量准确性（违反财务调节器规则）或延迟（违反服务协议）。 在本文中，我们提出了铁轨，这是一种容忍故障，弹性和分布式流媒体系统，为需要高负载和毫秒级别的潜伏期的场景提供实时滑动窗口。我们使用真实数据对轨道枪的初始原型进行了基准测试，显示出明显的延迟性，而较低的记忆使用情况下，独立于窗口大小。
Some mission critical systems, such as fraud detection, require accurate, real-time metrics over long time windows on applications that demand high throughputs and low latencies. As these applications need to run "forever", cope with large and spiky data loads, they further require to be run in a distributed setting. Unsurprisingly, we are unaware of any distributed streaming system that provides all those properties. Instead, existing systems take large simplifications, such as implementing sliding windows as a fixed set of partially overlapping windows, jeopardizing metric accuracy (violating financial regulator rules) or latency (breaching service agreements). In this paper, we propose Railgun, a fault-tolerant, elastic, and distributed streaming system supporting real-time sliding windows for scenarios requiring high loads and millisecond-level latencies. We benchmarked an initial prototype of Railgun using real data, showing significant lower latency than Flink, and low memory usage, independent of window size.
