当前位置:网站首页>Understanding and use of tp50, tp90 and tp99
Understanding and use of tp50, tp90 and tp99
2022-04-23 06:06:00 【New ape and horse】
One TP50、TP90、TP99 The concept of
1.1 What is? TP
TP yes Top Percentile Abbreviation , Chinese translation Percentile .
1.2 What is the percentile
Percentile is a statistical term .
If you sort a set of data from small to large , And calculate the corresponding cumulative percentile , Then the value of the data corresponding to a certain percentile is called the percentile of the percentile . Can be expressed as : A group of N individual Observed value Press The number Size arrangement . Such as , be in P% The value of the position is called P Percentiles .
1.3 TP50、TP90、TP99 How to understand
TP50、TP90、TP99 Is the engineering performance index , Take the time-consuming network request as an example :
- TP50: Indicates the minimum time required to satisfy 50 percent of network requests .
- TP90: Indicates the minimum time required to satisfy 90 percent of network requests .
- TP99: Indicates the minimum time required to satisfy 99 percent of network requests .
Two TP50、TP90、TP99 The calculation of
2.1 A simple example
for instance : There are four requests that take :
10ms,1000ms,100ms,2ms
Then we can calculate TP99:4 In this request ,99% The number of requests is 4*0.99, Round up, that is 4 Time , Meet all this 4 The minimum time consumption of this request is 1000ms, That is to say TP99 The answer is 1000ms.
2.2 python Realization
Application numpy Bag percentile Method realization TP50、TP90、TP99 The calculation of .
import numpy as np
data = []
with open(' File path ', 'r') as f:
for line in f.readlines():
data.append(float(line.strip()))
print(np.percentile(data, 95))
File path : It's the absolute path , Such as /User/local/a.txt, among a.txt Is the time-consuming array of network requests .
You can put... In the code 95 Modified into 50、90、99, To achieve TP50、TP90、TP99 The calculation of .
版权声明
本文为[New ape and horse]所创,转载请带上原文链接,感谢
https://yzsam.com/2022/04/202204220533487883.html
边栏推荐
- Create binary tree
- Get the value of state in effects in DVA
- Algèbre linéaire chapitre 2 - matrice et son fonctionnement
- 治疗TensorFlow后遗症——简单例子记录torch.utils.data.dataset.Dataset重写时的图片维度问题
- Reading of denoising papers - [cvpr2022] blind2blind: self supervised image denoising with visible blind spots
- container
- Chapter 3 of linear algebra - Elementary Transformation of matrix and system of linear equations
- What is the difference between the basic feasible solution and the basic feasible solution in linear programming?
- Remedy after postfix becomes a spam transit station
- Filebrowser realizes private network disk
猜你喜欢

Pytorch learning record (III): structure of neural network + using sequential and module to define the model

Pytorch学习记录(十三):循环神经网络((Recurrent Neural Network)

JDBC connection database

filebrowser实现私有网盘

容器

Development environment EAS login license modification
![去噪论文——[Noise2Void,CVPR19]Noise2Void-Learning Denoising from Single Noisy Images](/img/9d/487c77b5d25d3e37fb629164c804e2.png)
去噪论文——[Noise2Void,CVPR19]Noise2Void-Learning Denoising from Single Noisy Images

PyQy5学习(四):QAbstractButton+QRadioButton+QCheckBox

给yarn配置国内镜像加速器
![Reading of denoising papers - [cvpr2022] blind2blind: self supervised image denoising with visible blind spots](/img/fd/84df62c88fe90a73294886642036a0.png)
Reading of denoising papers - [cvpr2022] blind2blind: self supervised image denoising with visible blind spots
随机推荐
Multithreading and high concurrency (2) -- detailed explanation of synchronized usage
Denoising paper - [noise2void, cvpr19] noise2void learning denoising from single noise images
Pytorch学习记录(十三):循环神经网络((Recurrent Neural Network)
自動控制(韓敏版)
Solve the error: importerror: iprogress not found Please update jupyter and ipywidgets
Software architecture design - software architecture style
Pytorch学习记录(五):反向传播+基于梯度的优化器(SGD,Adagrad,RMSporp,Adam)
Pytorch learning record (III): structure of neural network + using sequential and module to define the model
线性代数第三章-矩阵的初等变换与线性方程组
JDBC operation transaction
Exception handling: grab and throw model
Anaconda
lambda expressions
Create binary tree
Latex快速入门
Pytorch notes - complete code for linear regression & manual or automatic calculation of gradient code comparison
String notes
Treatment of tensorflow sequelae - simple example record torch utils. data. dataset. Picture dimension problem when rewriting dataset
Class loading and classloader understanding
线性代数第一章-行列式