一个数据库或多个数据库

Question

So, today we had long conversation in company between project leader and programmers, shall we run one DB with all tables in it for our new project or to run multiple databases with each DB storing one set of module.

Project is about shop, we have separate (lets call it modules) such as users, payment methods, products, statistics.

Now one side said that we should place all of it inside one DB(its standard procedure) because it will be faster, with one query you can get all data, while other side said that we should split it between multiple databases so that its more secure, if someone breaches inside products, they wont see users tables as that DB will be on different server or virtual machine.

So my question is, what are pros and cons of having single DB for all data VS having multiple databases. I read few questions on stack-overflow, but none of them were precisely about pros and cons. And if having multiple databases is slower, how to speed it up?

Thanks!

Answer 1

Rather than providing you generic pros/cons (because everything depends on the use case...), I would say that we tend to prematurely optimize systems while optimization shouldn't be a problem in the future, if the whole system is architected with refactoring in mind.

IMHO, I had the same discussion some time ago and my conclusion is starting with a single database. It simplifies a lot of details:

• Single database to backup, less maintainance.
• You don't need to manage multiple connections.
• Multiple databases can break the chance to perform atomic transactions, a feature I would never throw away.
• You avoid synchronizing two or more databases to avoid integrity problems.

Also, since we're in the cloud computing era, infrastructures should scale horizontally. That is, if you need more power, add a replication node and distribute your load across multiple servers instead of scaling in the application level. This ensures your software will be still easy to maintain and develop, and good solutions should scale out easily if your code has quality and, of course, you've budget to support an increased load!