Android Question 10.000 users simultaneous

[Prosg]

Hello,

I have to developp a app for a big company who have a event.

What is the way to store data to have 5 000 - 10 000 users in the same times for 10 minutes ?

I think i must forget mysql ^^

Ty for your help

 

[eurojam]

a middleware technique like JMS, which builds a persistent queue and pulling and processing request one by one, could be one solution

 

[Erel]

MySQL is a very powerful database.

You can see some numbers here: https://www.b4x.com/android/forum/threads/server-performance-measures.37502/#content

 

[sorex]

it all depends on how you write the app and the back-end.

don't send multiple small data requests but group everything to 1 data push.
merge queries if possible to have less independent database calls
etc...

 

[KMatle]

Must have:

- good database design (Indexes, etc.) = 80% of the way
- as @sorex said, no multiple requests (like sending 100 httputils jobs for 100 rows). See here: https://www.b4x.com/android/forum/t...ring-via-job-download2-to-a-php-script.50919/

What kind of data does the system store/retrieve (and how much)?

 

[Informatix]

Hello,

I have to developp a app for a big company who have a event.

What is the way to store data to have 5 000 - 10 000 users in the same times for 10 minutes ?

I think i must forget mysql ^^

Ty for your help

It depends strongly on what the users will do with the database (they'll only read? they'll write? they'll access the same table? they'll have complex queries? etc.) and how many servers you have at your disposal. Supporting a large number of users that will write in the same table on an unique server is not a simple task.

 

[Prosg]

they'll only read? => no
they'll write? => yes
the same table? => yes

1 dedicated server ^^

I think can't do this with mysql

It's like the app ask a question and 10 000 users answers in the same times

 

[sorex]

mysql won't be the bottleneck, your connection would.

 

[Informatix]

I think can't do this with mysql

I don't know MySQL. I'm only used to Oracle and SQL Server because we handle millions of users. You should try PostGreSQL. It's a free DBMS that should be able to handle so many users.
Do you have something to emulate the connections and test the load ? e.g. LoadRunner
What is your strategy in case of a bottleneck (too many accesses at the same time)?

 

[Prosg]

millions of user ok but my problem is 10 000 user in the same time^^

a database is not 30 users on the same times ?

 

[Informatix]

millions of user ok but my problem is 10 000 user in the same time^^

Yes I understood. That's why I suggested PostgreSQL (with PGBouncer for the connection pool).

 

[Prosg]

how many connexion in the same time ?

 

[Informatix]

how many connexion in the same time ?

Do you know what a connection pooler is? Because it's not the maximum number of database connections that really matters but the support of connection pooling by the DBMS. No DBMS accepts thousands of direct connections at the same time on a single server. You have to use a connection pooler.

 

[Prosg]

it's like a cache but for database. With a pool connexion i can have unlimited connexion at the same time but i need something like memory ?

Oh we are french together i send u a pm

 

[Informatix]

it's like a cache but for database. With a pool connexion i can have unlimited connexion at the same time but i need something like memory ?

Oh we are french together i send u a pm

I did not install a database server since at least 7 or 8 years so I would say only stupid things about the requirements nowadays. I don't know what are the modern tools to ease your task or the recommended architecture of today. But things should not be completely different. You need to think seriously about the following things:
- hardware and monitoring (to ensure that your server is up and works as expected);
- transmission of data (via internet? how are the data sent to the server?);
- multiple connections (connection pooler?);
- many concurrent writes (which scenario when editing the same data? when max number is reached? etc.);
- database performance (tables and records organization, triggers, indexes, etc.);
- backup (you have only one server, so that would be stupid to lose everything with a crash).

 

[Prosg]

And is it possible with xml ? and no database

 

[Informatix]

And is it possible with xml ? and no database

I'm not sure to understand the initial request.