Answer the question
In order to leave comments, you need to log in
Server under heavy load, uptime?
Hello!
I want to ask such an interesting question to which I have not been able to find an answer for a long time.
I don't know much about administration, so I'm asking for help.
There is a server
specifications:
CPU: 2xXeon L5520
RAM: 24Gb
PORT: 1Gb / s
php: 7.2
MySQL is also used
Apache web server
OS: Debian 9.x The
question is how to configure the server so that it can withstand 1000 people online at the same time +
Drive test traffic server started to freeze when there were 470 people, and fell when there were 600 people.
Give some advice to make the server work stably 24/7 under load, thanks in advance!
Answer the question
In order to leave comments, you need to log in
nginx phpfpm
Well, you have to rewrite the code so that it holds the load
There is a server
specifications:
CPU: 2xXeon L5520
RAM: 24Gb
PORT: 1Gb / s
php: 7.2
MySQL is also used
Apache web server
OS: Debian 9.x The
question is how to configure the server so that it can withstand 1000 people online at the same time +
Drive test traffic server started to freeze when there were 470 people, and fell when there were 600 people.
Give some advice to make the server work stably 24/7 under load, thanks in advance!
Either contact a specialist, but a real one, not a pseudo one, or conduct an audit on your own and answer the questions:
- which component of the system (base, php, web server) slows down the most with 470 users
- what to do with it so that it does not slow down
further correct
and so on in a circle until it's good
you have several options, the first one is to speed up the processing of a separate request, see where it has a bottleneck, etc. (switch from interpreted languages ​​to compiled ones, already keep some data in RAM and not read them from somewhere), this can increase performance by dozens or even hundreds of times.
the second is to build parallelization of request processing, change the scheme for processing these requests from synchronous to asynchronous, adding individual heavy tasks to a separate queue that would NOT block the processing of simple user requests (this can also increase performance dozens of times)
In any case, this is the task of the developers of your service -\u003e just "configure" the server, it will not give much output, you can, for example, increase memory, change disks, put a 20-core processor instead of a 10-core one, and you won’t see any increase at all, just ZERO.
Didn't find what you were looking for?
Ask your questionAsk a Question
731 491 924 answers to any question