Re: another failed attempt at database independence
- From: "Charles Schultz" <sacrophyte@xxxxxxxxx>
- To: ricks12345@xxxxxxxxx
- Date: Fri, 9 May 2008 14:57:28 -0500
To look at things in a different light, has anyone seen database agnosticism
successfully achieve the goal it set out to do? From my observations,
database independence does not target performance (or if it does, in a
negative sort of way), but rather, shorter deployment cycles and "easier",
more portable code for developers.
Surely, if it is not achieving that goal, surely folks would stop using it.
Right? =)
On Wed, May 7, 2008 at 10:21 AM, Rick Ricky <ricks12345@xxxxxxxxx> wrote:
> The Defense Travel System (DTS) is attempting to move to database
> independence. Last I read a few years ago they spent $600 million on this
> application up to that point. I'm sure its alot higher now. Probably close
> to $1 billion or more. It basically handles all of the commercial travel for
> the US Department of Defense (over 3 million users). They have over 2 TBs of
> data. They did not design for archiving so it will grow indefinitely.
> They are currently working on a "technical refresh" (supposedly that is
> their PR word for "pay us to write this piece of junk software again"). They
> wrote their new modules against a mySQL database using an outsourced
> sub-contracting company(which made money even though this failed
> completely. I think the company is Dovel. Not sure. Might be IDC). They
> wanted to prove they could make the application database independent. They
> used a tool called Hybernate to generate all their queries. Probably spent
> millions of dollars on this re-write of the code.
>
> They deployed it to production 2 weeks ago and it was so bad that the whole
> system was down for 3.5 days. This means EVERY person who works for the
> department of defense could not book commercial travel
> or get reimbursed or book hotels or get reimbursed for taxis or meals, or
> CHANGE FLIGHTS if they were overseas for 3.5 days. They had to back out the
> changes. It totally failed. Now since this is a time and material
> contract(they make more money if they screw up), they are getting paid more
> money to fix it.
>
> They do not have any code built into their application to let them detect
> where the performance problems may be. Its so pathetic I have been told
> their DBAs laugh at the rest of the team in their meetings. More of my tax
> money down in flames. They already paid for the oracle licenses. Migrating 2
> TBs of data that is GROWING to another database is so unlikely it is
> laughable. Yet the DoD got sold on database independence. They are not
> allowed to use ANY oracle features. It would mean days of down time just to
> move the data to the new database and this is before even testing it. That
> is not going to happen. The data model has no normalization or primary keys
> at all (they ignore their DBAs).
>
> btw, if you google defense travel system you will see criticisms of the
> project going back almost 10 years. yes its been in constant development for
> 10 years and its still not done.
>
--
Charles Schultz
Other related posts:
