Re: schema help

On Dec 25, 12:37 pm, magawake <magaw..._at_gmail.com> wrote:
> Hello:
> I am trying to learn SQL and RDBMS theories. To acheive this, I
> decided to assign myself a project.
>
> My project is basically, account for my NFS mounts, and keep track of
> them. I want to see which filesystems are growing, and which
> filesystems are staying the same. I will get an inventory daily. I
> will grab the data and place it into a database, using PERL or AWK.
>
[some illustrative data]
>
> So my intentention is, to show do stats on these filesystems, and see
> when it grows and when it stays the same. I will get these stats
> daily, and hopefully in a month, I can see what is growing, shrinking,
> and staying the same.
>
> Any ideas on the schema? do I really need a RDBMS for this?
>
> TIA
A fascinating question I've long pondered. Since I am interested in the general question, I'll take the liberty to restate it, accurately I hope:

Since a relational database represents the model state of a known universe, is it possible to ask/answer the question: How has the database changed over time?

Snodgrass advocates adding begin/end time attributes, for the data entry window and the data validity window (with representations for minus- and plus infinity). But am I really modeling change-over-time or just adding timeframes to snapshots? (Like using a timed exposure to take a picture with a camera when there's too little light.)

Would it not be better to utilize the OP's data capture approach by periodically evaluating a query, saving the query response externally and computing "deltas" between query responses (extensions) over time?? (This is fundamentally the data warehouse approach in which a relational database is used to capture periodic snapshots.)

What do other people think about this? Is it better to implement a temporal database or to capture snapshots of a non-temporal database (externally or in a data warehouse) and compute the changes?

Rob Received on Tue Dec 25 2007 - 23:40:08 CET