I now have a website. I want to simulate how millions of users can access my website at the same time and configure the Server properly, what is hardware required? How can we change the number of users who access the service per second for tens of millions of users? I now have a website. I want to simulate how millions of users can access my website at the same time. How can I reasonably Configure the server and what hardware support do I need?
If tens of millions of users access 0.1 million users per second, how can this problem be solved?
Reply content:
I now have a website. I want to simulate how millions of users can access my website at the same time. How can I reasonably Configure the server and what hardware support do I need?
If tens of millions of users access 0.1 million users per second, how can this problem be solved?
Do you know that 10 million access per second means how much PV a day?
100000*3600*24 = 8640000000 PV
This is not counted as the peak value concurrency.
I will give you more information.
The daily pv of alexa ranked third youtube.com is close to the pv of your website.
Therefore, do you think that a person can construct the access level? Before asking a question, you 'd better know where you are.
Let me give you some dry articles to give you a conceptual understanding of the so-called high traffic.
AB, qps, server performance pressure
Ant becomes an elephant: Talking about how regular websites grow from small to small
First of all, as the landlord said above, don't think about simulating 0.1 million concurrency at once. If you want to learn the high concurrency architecture, I suggest you first use the linux webbench, jmeter, or AB stress testing tools to simulate concurrency, first from 100 per second, and then slowly go up, of course, several servers with better performance must be used for stress testing. Otherwise, if the performance of the sub-server is poor, even if the concurrency you set is higher, the number of servers that you set will not exceed 100 in one second, we recommend that you use several more stress testing servers to make them distributed, so that thousands of concurrent jobs can be performed per second. However, if you want to simulate 0.1 million concurrent jobs at the same time, unless you have N machines with good performance .. During the test, you will find that, after hundreds of concurrent requests per second, your application layer is under a great pressure on the database layer, you want to simulate 0.1 million of the concurrency at once. Even if the simulation is successful, all the layers, web terminals, application layers, and databases will be suspended .. Because the 0.1 million concurrent architecture won't be able to be handled by several machines.