Re: Scalability Design Questions

On 9/8/07, novnov <novnovice(at)gmail(dot)com> wrote:
>
> What happens when a database needs to scale beyond a single server? Is this
> something PostgreSQL can do?
>
> Further, does a database need to be designed differently if it's going to be
> used on more than one server? I know there are many issues that might effect
> a database's ability to scale; but specifically, will a well designed single
> server database be easily adapted to multiple servers, or will the design
> have to be adjusted?

There are so many ways in which an application may need to scale.
What if your data is mostly static, but has 100,000 users an hour or a
minute hitting it (think google, news sites). what if you've got to
handle 1,000 airline reservations a minute? What if you've got to
handle 500 inserts per second, but only have to aggregate the data at
the end of every hour and then throw it away?

The way you scale depends very much on what you're doing.

While the idea of just throwing more machines in a cluster that act as
one big machine seems simple, the actual implementation and software
to make that happen is not, and it's why Oracle can charge as much as
the national budget of a small island nation to implement a large RAC
cluster.

PostgreSQL can handle a lot of very large jobs, but it really doesn't
have anything as complex, sophisticated (and trouble pron I might add)
as oracle's RAC. Start cutting checks for $200,000 to a few
developers and you might start getting somewhere on it though. :)