当前位置:网站首页>Distributed High performance
Distributed High performance
2022-07-21 19:19:00 【Idle cat】
Measures
response time : The time when the system responds to the request . For example, the system deals with a HTTP Request needs 200ms, This 200ms Is the response time of the system .
The response time of a request is of no value , The factors that affect the time of a single request vary , Only a response time curve over a period of time can help check the problem , If you sort according to the request response time from small to large, you can more intuitively see the proportion of each request time , Sensitive to slow requests , This method becomes TP indicators ; If you sort by timeline , Help to check the problem .
Mean response time AVG: The most commonly used , such as 1 Ten thousand requests , among 9900 Next is 1ms,100 Next is 100ms, The average response time is 1.99ms, Although the average time spent has only increased 0.99ms, however 1% The response time for requests has increased 100 times . Can't reflect the slow request .
TP Quantile index : In a period of time , Count the time consumed by each call of the method , And sort these times from small to large , And the result is .TP34 Quantile mark No 34 Response time of requests .
The lower the index, the better :
- TP90 It is the minimum time required to meet 90% of network requests .
- TP99 Is the minimum time required to meet 99% of network requests .
- TP999 It is the minimum time required to meet 999 / 1000 network requests .
Empirical value
From a user experience perspective ,200 Millisecond is considered the first cut-off point , Users don't feel the delay
1 Second is the second cut-off point , Users can feel the delay , But it's acceptable .
therefore , For a healthy, highly concurrent system ,TP99 It should be controlled in 200 Within milliseconds ,TP999 perhaps TP9999 It should be controlled in 1 Within seconds .
Index setting
For highly concurrent systems , Performance targets are set with both throughput and response time in mind
For example, to say : Every second 1 Ten thousand requests , The average response time is controlled at 50ms following ,TP99(99% Request ) Control in 100ms following .
END
边栏推荐
猜你喜欢
奈雪的茶布局元宇宙 传统品牌如何转战元宇宙营销
外设驱动库开发笔记44:DDC114 ADC驱动
Web APIs DOM- 网页特效篇-元素大小和位置
Configuration du Service FTP
TiDB 分布式批量解决方案
kubernetes的这几种存储卷,别再傻傻分不清了
JMeter学习笔记003-JMeter之CSV参数化
LogMAP: logic based and scalable ontology matching
Translation and interpretation of the paper: anytime bottom up rule learning for knowledge graph completion [anyburr]
[information retrieval] implementation of information retrieval system
随机推荐
Filter listener
分布式.负载均衡
Software test interview question: in your previous work, what did a software defect (or bug) record contain? How to submit high-quality software bug records?
软件测试面试题:详细的描述一个测试活动完整的过程。(供参考,本答案主要是瀑布模型的做法)
[QNX hypervisor 2.2 user manual]8.6 interrupt
5. Customize global AuthenticationManager
05 正则表达式语法
分布式.高性能
Software testing interview question: what is the strategy of software testing?
[experience sharing] mathematical modeling paper format requirements and summary of common problems
分布式.高可用 高可扩展 指标
ThreadLocal
手把手教你安装MySQL数据库
Session 和 Cookies 有什么区别?
PyTorch 深度学习实践 第2讲/作业(Linear Model)
论文翻译解读:Logmap:Logic-based and scalable ontology matching
[QT入门篇]定时器QTimer类
Spark RDD, application case of spark SQL
软件测试面试题:软件测试各个阶段通常完成什么工作?各个阶段的结果文件是什么?包括什么内容?
How to realize copyright protection of Digital Collections