JMeter stress test index interpretation

RT(response time)：

What is? RT?

RT It refers to the time span when the system receives the request and makes the corresponding request

But it's worth mentioning RT The higher the value of , It doesn't really mean that the throughput of our system is very high , for instance , If there is packet loss when data is transmitted in the network , Less data to transmit , The speed will increase , But it's an accident , Not that our system throughput is high

Besides , There are different standards for different systems , For a game software , Response time is also known as delay . stay 100ms within , The experience of the game will be very good , Soar to a few seconds , Just stop playing , But for a compiled system , Compiling a set of source code can take tens of minutes , People are willing to wait java In the application GC It also leads to the overall system RT soaring

---

Throughput throughput ：

What is the amount of vomit ?

Baidu Encyclopedia : The generalized throughput is : Number of successful transfers per unit time , In the results of our stress tests , Throughput can be understood as the number of requests successfully processed by the system in unit time

Throughput of a system and request Yes CPU Consumption of , External interfaces and IO And so on ,

For example, a web How beautiful the application code is , But every request will go to the database , When the concurrency is up , The database will take up a lot of CPU load , Systematic IO soaring , It may even cause other software to have to wait mysql You can't get to CPU Time slice of . The bottleneck of the system backlog to mysql here , This web The throughput of the application must not be very high

Besides , We usually describe throughput in this way : The concurrency number is xxx when , The vomit and swallowing capacity of the system is yyy

---

Number of concurrent users ：

It refers to what the system can carry , The number of users who can use the website at the same time , This metric seems to reflect the performance of the system more intuitively than throughput , But often users in the system have different behaviors , For example, unregistered , online , Send requests at the same time, etc , In short , Consider using online users and the number of users sending requests at the same time as performance indicators , It's more intuitive to treat online users as performance indicators , It is more accurate to regard the number of users who send requests at the same time as the performance index

---

QPS (query per seconds)：

Query rate per second , It is used to weigh the number of traffic processed by the server within the specified time
Calculation formula : QPS = req / sec , The average number of requests per second

---

TPS (transition per seconds)：

TPS (transaction per second) Represents the number of transactions executed per second , It can be calculated based on the number of transactions completed in the test cycle . for example , The user executes... Every minute 6 One transaction ,TPS by 6 / 60s = 0.10 TPS. At the same time, we will know the response time of the transaction ( Or beat ), Take this example ,60 Seconds to complete 6 Transactions also represent that the response time or beat of each transaction is 10 second .

---

PV and UV：

PV Traffic volume (Page View) Every time you open a page or refresh a page pv+1, It reflects the number of times a page of a website has been visited
　　UV Number of visits (Unique Visitor) namely , Number of visits by independent visitors , In other words, a computer is a visitor , adopt QPS Estimate PV
　　 The formula 1 :
　　　　QPS = total pv Count 0.8 / Seconds per day 0.2
　　 The formula 2:
　　　　pv = QPS36008
　　 according to QPS,PV Estimate the number of servers
　　 Number of servers = Every day's total PV / Single server PV
　　 principle : Every day 80% Our visit focuses on the current 20% Time period , this 20% The time of is called peak time peak period
　　pv = total pv Count 0.8 / Seconds per day 0.2

---

Aggregation report ：

Parameter description ： The unit is millisecond

Label： Definition HTTP Name of the request 
Samples： Indicates how many requests were made in this test 
Average： Average response time —— By default, it's a single request The average response time of 
Median： Median , That is to say 50% Response time of the user 
90% Line：90% Response time of the user 
Min： Minimum response time to access a page 
Max： Maximum response time to visit a page 
Error%： Number of error requests / Total number of requests 
Throughput： By default, this represents the number of requests completed per second （request per second）
KB/Sec： The amount of data received from the server per second 

---