See "What you need to know to do performance testing"

See "What you need to know to do performance testing"
Need to understand the best number of users and maximum number of users
Before you start a performance test, you need to know:
1, the customer's request
If the system meets * * User logon, average per user logon time is not more than * * seconds
2, the current system performance
The maximum number of users and the optimal number of users, the system analysis and find the pressure point
3. System bottleneck
Analyze logical business that can cause bottlenecks to your system
4. Performance status under long-time pressure (strength test)-"Memory Overflow"

Test environment
1. Hardware
Server, CPU, memory
2. Software
Front end, middleware, server
3. Network environment

Find the pressure points of the system

What you need to know after doing a performance test-"get effective" performance requirements
Example: "Actual demand" for a business in the securities industry
Total system capacity Dari commissioned 60 million, sold 90 million pen
System processing speed 7300 strokes per second, peak per second, 10000 strokes
Actual shareholder account Number 30 million
-"Demand: Best Concurrent users demand: 7300 strokes per second
Maximum Concurrent users demand: 10000 strokes per second
Base data capacity: Actual shareholder account number 30 million
Business Data capacity: Day care 60 million, deal 90 million
If we are based on the following 2 assumptions:
1. With a 2/8 ratio, that is 80% of the business at 20% Peak time, 20% of the business in the non-peak time of 80% to complete, then we can get peak processing business volume 150 million is 80%, non-peak processing traffic 120 million 150 million is 20%;
2. The 1-day system runs for 20 hours and 4 hours for non-business background processing, then 20% hours for the peak time of 20 hours, and 4 hours for non-peak hours 20.80% hours.
We can calculate to:
1. Average peak processing speed of 120 million/4*3600 seconds close to 9,000/s;
2. Average non-peak processing speed of 30 million/16*3600 seconds approx. 500/s;
Given the special circumstances, we recommend that the actual peak processing speed be up to 1.5 to twice times the average peak processing speed of the theoretical calculation, so the recommended peak processing speed for the final determination is 9,000 *2=18000/sec. We explain to our clients that the original requirements are likely to be handled effectively in the event of a particular situation, and the client finally accepts our claims and adjusts their needs.
This is called demand development, through the analysis of the customer's unreasonable needs, to meet their most fundamental needs "total system capacity to meet the day commissioned 60 million, deal 90 million", and processing speed is their own needs to estimate, not accurate.
The so-called demand development, that is, to eradicate the core needs of customers, for customers to design a complete demand system, and even according to customer business development needs, for customers to design core needs and demand system.

Taken from: http://www.cnblogs.com/jackei/archive/2006/12/12/589473.html

Common metrics for performance testing
Workloads = virtual users
Response time (speed of processing requests)
Throughput
Ti-the throughput of the request packet from the client to the server
to-"Server returns the throughput of the request packet
Page Hits/Requests
Response/Successful response
Hits per second: Server processing power

This article is from the "9024359" blog, please be sure to keep this source http://9034359.blog.51cto.com/9024359/1719133

See "What you need to know to do performance testing"