知识卡片
延迟按阶段预算与背压
把端到端响应时间分配到流水线各阶段,并监控队列长度,能看出时间在哪耗尽。可并行化独立工作、减少I/O、用背压限制新请求;但阶段有依赖时并行收益有限,异步提前响应也不等于后台业务完成。延迟优化应围绕用户所等待的结果定义终点。
相关:[[延迟预算约束接纳速率]]。
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。
在整书笔记中的位置
原书出处 · 1 处
- 理解性能的心智模型 · 查看原书依据
For example, if a service call must finish within 100ms, we can budget each part of the code to respond within a given time period. In general, we can break the system into multiple parts in the pipeline and specify the latency limits for each part while monitoring the queue lengths.
参考来源
- 位置:《Software Architecture and Decision-Making》第3章《理解性能的心智模型》(来源项目:book-sources;源文件:extracted/Software Architecture and Decision-Making Leveraging Leadership, Technology, and Product Management to Build Great Products… (Srinath Perera)/_epub-src/ch03.xhtml)
- 结论依据:原文的定义、比较或例子支持“把端到端响应时间分配到流水线各阶段,并监控队列长度,能看出时间在哪耗尽”;正文据此说明用途和边界。
- 原始内容:For example, if a service call must finish within 100ms, we can budget each part of the code to respond within a given time period. In general, we can break the system into multiple parts in the pipeline and specify the latency limits for each part while monitoring the queue lengths.