Haha~ This problem is inherently problematic. Just like you are asking, "There are 100 people sleeping on the desks at the same time in our school. How many people are there in our school?" It is a truth. Even if someone can figure out how many students there are in the school, that is only a rough figure.
.
However, because ordinary children's shoes don't have much concept of 100 concurrency and 1000 concurrency, customers are often asked: "How many users can you say 1000 concurrency can carry?" In order to solve this nonsensical problem, use it today. Probability to answer this question.
If the user performance requirements are as follows: Support 1 million registered users Performance requirements analysis:
1. According to user requirements, this system must support 1 million users. What is the configuration of the performance machine? What is the peak? bandwidth? Wait
2. If the company's test environment is used, what kind of performance should be used this time? Performance evaluation, load test
, Strength test?
3. How to calculate the number of concurrent users? Response time? The performance index is determined: Because the user's performance requirements are too wide, no specific value has been set.
So how do I carry out subsequent work?
1. make sure to use the company's test environment without considering environmental issues. In other words, the client
A system such as, server and bandwidth can be ignored. This is fixed.
2. Consider whether the performance of the system previously developed by this Project team can be used as a reference value. Solution: Find out that the project team has used two projects concurrently, and the performance indicaTors are weighted. The browsing function: the number of concurrent sessions is 1100, the average response time is 363 seconds; the average response time per user is 0.33 seconds. 3 concurrent users per second. One system has 5 million users, and the other has 3.2 million users. And the second system has been running normally, but the current second system has 3 servers each. It can be concluded that a server is 1.66 million or more. (Because there is a power request in the server)
3. 1 million users, then how to calculate the number of peak active users per hour
? Solution: Using the 80•20 principle to calculate the number of peak active users per hour is 66.67 million/hour; then the number of concurrent clicks on the same function point per second should be 18.5. 4. How to get the number of concurrency? Solution: How many function points does this system have? There are 153 function points; that is, the system is at its peak
The Shiyi function will be clicked 1258 times, 0.35 times per second. (Do not consider the interval time) Consider the previous value of this project group. Initially set the number of concurrency to 1100, mainly based on the browsing function, followed by query and new addition.
5. What kind of performance should be tested. After repeated consideration, all three performances should be tested. Execution performance
: Evaluation, according to the third point in the determination of performance indicators, set the user's concurrency to 300-350, depending on the situation. Load test, with 1100 as the starting point, intensity test, which is 15 hours and 24 hours. Performance test results: It is found that the maximum user support of this system is 1100. The maximum number of failed users is 209, and the response time is 315.
According to the above situation, it can be concluded that: When 1100 users are concurrent, the total response time of users is 315 seconds (that is, the average response time per user is 0.005 seconds), of which 209 failed users are at most, but successful users can basically complete subsequent Operations. The maximum number of stable users required by the current system. It can be concluded that the maximum number of concurrent users supported by the system in the newly added function points is 1100. According to the ratio of 1*100, the number of peak active users per hour is calculated to be 110,000/hour; the 80•20 principle is used to calculate that the number of registered users supported by this system is approximately 1.65 million. The performance requirements of this system support 1 million registered users on a large scale.
SO, the maximum concurrent number of the system is about 1100. In other words, this system can support about 1.5 million users on one server. Is this number beyond your imagination? Allow the editor to make a bold prediction: At present, the daily concurrent traffic of the website is less than or equal to 1, which may account for about 90% of the total number of websites.
哈哈~这个问题本身就有问题。就像你在问“我们学校上课同时有100人趴在课桌上睡觉，请问我们学校有多少人？”是一个道理。即使有人能算出全校有多少个学生，那也只不过是个大概数字
。
但是由于一般童鞋对100并发、1000并发没有太多概念，因此呢，还有经常被客户质问：“你说的1000并发到底能承载多少个用户呀？”为了解决这个无厘头的难题，今天就用概率来解答一下这个问题。
假如用户性能要求如下：支持100万注册用户 性能需求分析：
1、根据用户的要求，本系统要支持100万用户，其中性能机器配置如何？高峰值是多少？带宽？等
2、如果都是采用公司的测试环境，那么本次性能应该做哪几种性能？性能评测、负载测试
、强度测试？
3、怎么算出并发用户数？响应时间？ 性能指标确定： 因为用户的性能需求太广，没有定到具体的数值。
那么我怎么开展后继的工作？
1、确定采用公司测试环境，不用考虑环境问题。也就是说，客户端
、服务端以及带宽等一系统都可以不用考虑，这是固定。
2、考虑此项目组以前开发过的系统性能情况，能否做为一个参考值。 解决方案：找出本项目组以并发过二个项目，其性能个项指标进行求权。其中浏览功能：并发数为1100，平均响应时间363秒；每用户平均响应时间为0.33秒。每秒中并发3个用户。其中一系统用户已达500万，另一系统用户为320万。并且二系统一直运行正常，但目前的二系统的服务器各为3台。可以得出一台服务器为载166万，甚至更多。（因为服务器中有求权的关系）
3、100万用户，那么怎么计算出他的每小时峰值活动用户数
？ 解决方案：采用80•20原则计算得到每小时峰值活动用户数 6.667万/小时；那么每秒中的同一功能点点击并发数应该是18.5。 4、怎么得其并发数？ 解决方案：本系统有多少个功能点？功能点为153个；也就是本系统在高峰值
时一功能将被点击1258次，每秒点击0.35次。（不考虑间隔时间）考虑以前本项目组的数值。初步设置并发数为1100，主要以浏览功能为主、其次是查询和新增。
5、应该测试那种性能类型经再三考虑，三种性能都进行测试。 执行性能
： 评测，依据性能指标确定中的第三点，将用户的并发设置为300-350，看其情况。负载测试，以1100为起点强度测试，为15小时和24小时为准 性能测试结果： 发现本系统最大用户支持为1100.失败用户最高为209，响应时间为315。
根据上述情况，可以得出： 1100用户并发时，用户一共响应时间为315秒（即每用户平均响应时间0.005秒），其中最高产生209个失败用户，但成功用户基本上可以完成后续操作，符合现系统要求的最大稳定用户数。由此可得出本系统在新增功能点中支持最大用户并发数为1100。按照1*100比例，计算得到每小时峰值活动用户数11万/小时；采用80•20原则计算得出本系统支持注册用户数约为165万。而本系统性能需求大规模支持100万注册用户。
SO，系统最大并发数为1100左右。也就说此系统在一台服务器上可支持150万左右的用户数。 这个数字是不是超乎你的想象，允许小编大胆的预测一下：目前网站每天的并发流量小于、等于1的可能占到总网站数量的90%左右。
- 内容分页 1 2
