D
D
das_hund2015-02-11 16:44:18
Highload
das_hund, 2015-02-11 16:44:18

Where can I read a lot about a real project with balancing and distributed load?

Hello,
there is a website by Ivan Blinkov and similar ones are briefly described there on VKontakte, etc. their architecture.
There are short articles like this one www.aosabook.org/en/distsys.html
There are also great tutorials on specific systems like - Redis, Mongodb, Hadoop, Mysql optimization etc.
But.
There is no description of a specific real highload project anywhere. I'm not interested in Facebook or flickr.
More real small examples interest.
But the breakdown of the topic is as follows:
server architecture - real balancer schemes, servers, how requests go where, how carousel balancing algorithms are applied, how the server hierarchy is built, what systems are configured - nfs is there or something else. Dns - balancing. Data storage is distributed.
Any fpm fork, etc.
All in a complex on a practical example.
There are general articles again, like on the Bitrix site - the scheme of communicating with the nginx proxy gives statics behind it the server and database replication. But this is too general again.
There are specialists who do this every day and can give a course - real books in English or Russian. Surely there are few such books in Russian.
One man gave me a few years ago, but I lost the link and connection with him. He worked in some homemade highload project. But he is no longer on the horizon.
If anyone can give some good advice, please. But please do not write general things like, get into a highload project and learn or study windows first. Read Tanenbaum.

Answer the question

In order to leave comments, you need to log in

4 answer(s)
P
Puma Thailand, 2015-02-11
@opium

So you think in vain that something common is written on the Bitrix, it’s just written there approximately the way everyone does it.
99.99% of projects on the Internet do not use any tricky schemes.
in general, the most common scheme is always
at the front any balancer more often nginx less often light or ha proxy
database most often master slave replication for splitting records and reading from two server
servers application most often some kind of php-fpm for php or uwsgi for python.

0
0ct0g3n, 2015-02-11
@0ct0g3n

First you need to throw away Bitrix

S
sevmax, 2015-02-11
@sevmax

When building a project, make sure that crashing a few random servers does not drop the entire project.
If you use a balancer, then you need to be able to transfer IP to a spare if the main one "left". If you are building databases, then you should automatically select a new master if the old one has gone down, and the fall of the slave should not affect performance in any way.
Don't use one server to run critical tasks, use redundancy wherever possible.
Well, make sure that the hoster has duplicate servers located on different subnets so that the service continues to work in case of problems in the hoster's internal network.
Good luck to you!

S
SilverSlice, 2015-02-12
@SilverSlice

Not a real project, but still a lot of practical information: ruhighload.com

Didn't find what you were looking for?

Ask your question

Ask a Question

731 491 924 answers to any question