re: interview question on schema design

From: Dba DBA <oracledbaquestions_at_gmail.com>
Date: Mon, 10 Dec 2012 15:24:30 -0500
Message-ID: <CAE-dsO+VY_6wPjngZcgpTkhvU0EvQvNkFysbS7tvkW=QH2osJQ_at_mail.gmail.com>



About 2 years ago I was working with a java architect who hated relational databases. He was convinced that a relational database could not handle the recursive style friend relationships that facebook had. I didn't get a chance to explain to him that yeah you can and that I think facebook primarily uses relational databases. I think they have some other things they use also. Then again this guy also told me that views were evil and his developers wanted to write their own database. this was a system that was about 100 gbs. 98 gbs of it were in a couple of CLOB fields. They were stored .pdfs. So you had 2 gbs of relational data and it was really slow. I turned on oracle auditing (just the basic stuff) and audit all the sqls to the database. Locked in and clicked a few buttons. Each button click was 15-25 separate sql statements. I tried to explain to him that if you cut down the number of sqls and did some joins. possibly simplified the table struct, this would run really fast. Got a blank look. All the SQL was generated. They considered the DB the 'persistence layer'. There were alot of 1 to 1 relationships mapping to objects as well. This lead to even queries.

For some reason its only the java people who are like this. I never run into this problem with C, python, .net or anything else developers.

On Fri, Dec 7, 2012 at 4:25 PM, kyle Hailey <kylelf_at_gmail.com> wrote:

> Here is an interview question using facebook for an example.
> It's a data modeling question that applies to a lot of social network sites
> and startups:
> http://dboptimizer.com/2012/11/07/facebook-news-feed-performance/
>
> I haven't answered the question, just provided some ideas.
>
> - Kyle
>
>
> --
> http://www.freelists.org/webpage/oracle-l
>
>
>

--
http://www.freelists.org/webpage/oracle-l
Received on Mon Dec 10 2012 - 21:24:30 CET

Original text of this message