[Asterisk-Users] OT: Best DB

Mohit Muthanna mohit.muthanna at gmail.com
Tue Mar 15 17:34:18 MST 2005


On Tue, 15 Mar 2005 18:52:00 -0500, Giudice, Salvatore
<Salvatore.Giudice at fmr.com> wrote:
> So, let me see if I am right. You run a support shop? You want your
> database to validate your data for you instead of leaving that logic to
> your application? Usually, a database is considered to be an asset worth
> protecting from unvalidated user input. Also, do you routinely try to

That is the silliest thing I've heard yet. I'm hardly a database
designer and even I know that there are many reasons for "server-side"
validation.

Say you have different client interfaces (command-line, web, RPC etc.)
for your app; in this case, why implement and maintain validation
logic in the clients. Sure, you can use a three-tier architecture and
have the middleware do the validation; but this is many-times not
practical.

What if your client app (or middleware) has bugs? A simple
buffer-overflow attack on a client or middleware piece of code can
potentially render it's validation moot.  Server-side validation can
atleast add another layer of security to catch errors in higher
layers. One could then decide (wieghing the tradeoffs), how much (or
how little) client side validation to add.

What about data / referential integrety? You expect the application to
take care of all that?

It's probably not a good idea to lecture someone on database design
practices when you have a few lessons to learn yourself.

> I agree that postgreSQL and MySQL have different feature sets. Your
> application design may drive your selection based on feature
> requirements.

And the general consensus here is that MySQL is _not_ the ideal
solution for heavily loaded applications or large datasets.

> with your observations. The largest of my past applications involved a
> ridiculously high number of batch/blind inserts and periodic data
> condensation with replicated storage for high level report optimization.
> I ran this app using a Beowulf cluster for parsing and two 8-way cpu
> servers running MySQL with a 2-terrabyte ultra160 storage array. I
> realize this is not the typical user experience, but I can tell you that
> we were able to handle a peak of 700k inserts per hour. MySQL gave us
> very few problems and probably had a cumulative downtime of
> approximately 4 days per year until the project was decommissioned. When

Right... 700k/hr inserts of "security events" from your IDS. Parsing
on a Beowolf cluster? 8-way cpu servers. Sounds like fun.

Mohit.


-- 
Mohit Muthanna [mohit (at) muthanna (uhuh) com]
"There are 10 types of people. Those who understand binary, and those
who don't."



More information about the asterisk-users mailing list