Scaling out

[BlaineSch]

I have been reading on "scaling out". I am trying to get a grasp on it still.

Basically from what I can tell, scaling out would get it so you have multiple servers with the same data, one server has writing ability the others have read abilities, the ones with read would continually gather data from the write server. Right so far?

Depending on the data and what you are using to find it, this does not sound like a good idea to me. For one you have 2+ servers with identical data. Why not find a way to break the data up into two different servers altogether, like breaking up the tables into different severs, if a few of the tables are the most high demand break it up into multiple servers based on a variable like if their last name is a-m go in the first, or n-z go in the second, obviously those should be adjusted accordingly and may make it difficult to search for unless you knew exactly how you were planning on finding it.

But let's say you have a user table with just a id, username, password, and email.

Then we have two tables both identical on separate servers, you know the ID of the user, but the ID of the data is one of those dynamic ones, you can easily tell if the number is even or odd ($i%2==0), the even ones would be the first, the odd the second. Then you can select all 20 rows relating to the user data.

Logical? Better ideas?

[BlaineSch]

Come on guys help me out here!

[debtboy]

I could be wrong, but that's not exactly how I understand Scale-out.
Let's say we have 2 DB servers each with an identical table
horizontally partitioned. One server would, as you say, contain data for last
names A-M while the other server would contain the same data for
names N-Z.
The data is combined by creating a view which the final query can draw from.

That's how I understand it.

I've never worked on a scaled-out system, but I have spent a little time
with a data warehouse/webhouse which did contain duplicate data
from multiple databases, but as you would expect, that system was dropped.

[BlaineSch]

Duplicate data just seems like I am wasting space. I realize space is cheap but when your talking 5million rows... thats wasteful!