[thelist] Heavy-Duty Databasing?

PJ -

Couple comment here..

Palyne wrote:

> There could be (in dreamland, the highest estimate) about 10,000
> different items on the site that will require a stored response.
> Times another dreamland maximum of 3000 students, that puts my
> max design around 31,000,000 answers to database. I keep
> thinking there must be something wrong and I have WAY too many
> zeroes, but...

31million might seem like a lot at first, but most internet sites are
way over this, so don't let it scare you..

> When I consider ways to break this up for table design on the
> storage of answers part, I realize that no matter how I look at it, it's
> huge.  We're not an intranet, we're not on shared server load, and I
> can do more than one domain (e.g., canonicals) but at least with
> my current hosting provider I can't do any more than one one
> datasource per domain, and so, it seems like it would really
> complicate a project that requires such tracking and integrated
> data.

Not sure if I follow you correctly here.. Are you planning on having
your DB hosted? You should look into co-location..

> 1.  How many tables can SQL server comfortably handle at one
> time?

As many as you need really.. Most any DB can handle a large number of
tables, this isn't an issue really..
 
> 2.  How many records can SQL server comfortably handle in one
> table?

Again, as many as you need(within reason). What you're really worried
about is performance, and if you use your indexes properly, this
shouldn't be an issue. You need to be concerned about how the number of
users connected to your DB is going to be affecting your performance
rather than the number of records.
 
> 3.  Would it be better for me to have tables where there are half as
> many records as there could be, but which are twice as long each?
>  (It would, right?)

umm ????

> 4.  If this were you and you had to do this project publicly hosted
> (not on your own hardware), what approach would you take to
> tabling this much info?  I assume someone on this list has done it.

If at all possible, you should look at getting a dedicated server for
this type of project, as others have mentioned :)

In summary, most any RDBMS can handle the kind of data you're talking
about here.. What it comes down to is the underlying structure of your
setup. Example: A 500Mhz NT box with 512MB of RAM would *NOT* be a good
choice for what you're looking to do here. Even though that sounds like
a lot of horsepower, you'd get problems due to certain shortcomings of
NT as an OS in terms of I/O performance, memory managment, and
stability.(Not that I'm trying to start a religous war, its just a
point)

I doubt many hosting providers can give you the kind of performance
you're looking for out of this project, which is why you should look at
co-location or a dedicated server.

Good luck at any rate :)

.djc.