Fog Creek Software
Discussion Board




K

Larry Ellison was predicting that Oracle 9i was the last database. Although he was lying (as usual, because Oracle will soon release or has released(?) 10g), his statement has a lot of truth to it.

The fact remains that Oracle 8i or SqlServer 2000 were the last database that **most** companies would require. I am not taking about monstrous databases needs likethose of amazon and ebay, but a majority of mid size to small companies. Oracle 8i or SQLServer2000 can easily handle upto 100 gigabytes(far more, actually) of information without too much fuss. Later on large transaction tables can be archived once the database reaches its limit.

What we may soon see is the death of database development. No, i am not saying Oracle or Microsoft will die. What i am saying is that a majority of companies will not easily upgrade from 9i to 10g because the incremental benefit accruing will be offset by the pain of cost+upgrade+laziness.

I have seen companies happily using 8i, even version 7 !.
So if 100 people upgraded from 7 to 8i, only 50 may do from 8i to 9i and 40 may do so from 9 to 10.

Of course, people using 7 may decide to switch to 10 and so on. But the "huge mass" of upgrades will be a thing of the past.  In the future (it may have already started now), people will only upgrade if not upgrading directly affects their business. For example, if microsoft releases a new version of windows and your existing database does not work with it, you may upgrade.

So what is the end result of it all?. The job of developing will turn into a job of maintaining. I suspect majority of these codebases will be sent to other cheaper countries like Eastern Europe, India for maintenance. The "core job"
of predicting future trends, deciding what needs to be developed will continue to stay in the U.S.  New product innovations in the DB world (like data warehousing or mining) will decrease to near zero.  Things may radically change once again only if there is some revolutionary development elsewhere in the hardware world -like some chip which can process at speeds fo 10 Ghz or a small hardware the size of your palm which can hold all your companys information. Then, existing databases may have to be rewritten to take advantage of it. But if incremental change (increase in processor speed from X to 1.2X) like what we are seeing continues, database development/innovation will be very very slow.

K
Tuesday, January 06, 2004

Ahh. My topic title was

"Has the last database been written?".
Can the moderator change it please?.

K
Tuesday, January 06, 2004

I don't think MSSQL 2000 is the last SQL that mid-sized companies will care about. The changes coming in Yukon can have a pretty significant impact on the way and types of application development you may choose to do.

Not sure about Oracle. Can't stand their tools, can't stand their CEO. I try to stay away from it.

Brad Wilson (dotnetguy.techieswithcats.com)
Tuesday, January 06, 2004

<<I don't think MSSQL 2000 is the last SQL that mid-sized companies will care about. The changes coming in Yukon can have a pretty significant impact on the way and types of application development you may choose to do.
>>

I too was under the same assumption. But people in JOS and elsewhere see things from a developers perspective. Business people see it in a totally diffferent way. To put it simply, you cant convince them to try out your new porsche if their 1980 lincoln works great.

K
Tuesday, January 06, 2004

"Has the last database been written?".

K,

You may want to add this statement to the Famous Last Words:
http://web.mit.edu/randy/www/words.html

Cletus
Tuesday, January 06, 2004

I'm not talking about necessarily upgrading existing databases. I'm talking about opening up whole new kinds of application development, by virtue of having this new tool available.

It's easy to justify in pure dollar terms. $5000 for the database, vs. longer developer time (which also means longer time to deployment). $5000 is generally about a week or two of development time, hardly any time to get anything done at all.

Yes, I do think it could have so large an impact in some cases that it could shave many man-months off a project. Against that, a new database license seems like a no brainer.

Brad Wilson (dotnetguy.techieswithcats.com)
Tuesday, January 06, 2004

And I thought you were referring to the K language and associated database:

http://www.kx.com/product_info/time-series-analysis.htm

On topic: there is also the issue of support to think about.  If the vendor retires support for a product do you still want to run your business on it?

What about 64-bit? For example: Microsoft aren't going to release older versions of SQL server with 64 bit support.  If your data/userbase grows and you want to migrate you will need to upgrade too.

Rob Walker
Tuesday, January 06, 2004

Database vendors force upgrades by dropping support of older products.

The old product might be fine, but who wants to be stuck with an unsupported product?

XYZZY
Tuesday, January 06, 2004

>The fact remains that Oracle 8i or SqlServer 2000 were the last database that **most** companies would require.

Yes and no.

Just think of the painfulness of dealing with RDBMSes, I don't know if Oracle and Microsoft (DB department) is responsible for removing this pain, but I am pretty darn sure my kids won't want to deal with the same crap we go through. The problem is that we have put up with the technology for 60 years, so it's a tough call. I think there are a lot of areas the Oracle and Microsoft database can grow, XML+web services, integrating analysis and decent ETL, giving Java/.Net a crack at running in-memory,  are all just a hint of the many things that must be implemented to make live bearable.

I think that both databases should strive to support in-memory databases and more object database support, as well as making the distribution of jobs across a farm even less painful. The dataset people work with now days are humongus. There's plenty for db folks to think about.

Li-fan Chen
Tuesday, January 06, 2004

See, the fallicy to your argument is assuming that all of the computer market is determined by one product -- the shrink-wrapped large-scale database.

The reality is that if we have, in fact, run out of stuff to do with the shrink-wrapped large-scale database, it will cease to matter as a differentiating factor and will end up being either free, open-source, widely licensed, or something like that.

I mean, TCP/IP stacks, extended memory managers, DOS extenders, etc. used to be a seperate product.  There used to be many many different 3D APIs (Anybody remember PHIGS?) 2D performance of a graphics card used to matter.  Stuff changes and different things become important.

Flamebait Sr.
Tuesday, January 06, 2004

Within a year of Yukon's release, will a majority of people rush to upgrade their SQL Server2000 applications to Yukon?  No.

Within a year of Yukon's release, will a majority of new SQL Server applications be written to run on Yukon?  Yes.

Obviously a lot of legacy apps will never be upgraded, but obviously a lot of new apps will use the latest version.  I mean... duh.

John Rose
Tuesday, January 06, 2004

> The changes coming in Yukon can have a pretty significant impact on the way and types of application development you may choose to do.

I think this actually reinforces the original poster's point: the database companies are pretty desperate for something new to talk about so that they can keep selling.

Also because they see MySQL coming up in their rear view mirrors.

Cool topic title, BTW! :)

Portabella
Tuesday, January 06, 2004

==>Anybody remember PHIGS?

I loved it!

If I remember correctly, Programmer's Heirarchical Interactive Graphics System.

Used it about 10 years ago (1993(ish)) in school, however never wrote a production app that used it.

Those were the days...

Sgt. Sausage
Tuesday, January 06, 2004

One trend I think will continue is improvement in data mining tools.  I was playing around w/ MS's the other day and was pretty impressed considering I'd never really heard of them before. 

Lee
Tuesday, January 06, 2004

----"The problem is that we have put up with the technology for 60 years, so it's a tough call. "----

What technology? Relational databases have only been common since the beginning of the eighties when hardware got advanced enough to run them.

And if you think they are painful to deal with wait for "Retunr to the 70's" when managers hyped on XML force everybody to go back to hierarchical databases, and a couple of years later start to understand why comapanies joyfuly spent hundreds of million dollars in the early eighties to dump hierarchical databases for RDBMSes.

Stephen Jones
Tuesday, January 06, 2004

Stephen, I'd just be happy if people would use the right tool for the right job...

Flamebait Sr.
Tuesday, January 06, 2004

"And if you think they are painful to deal with wait for "Retunr to the 70's" when managers hyped on XML force everybody to go back to hierarchical databases, and a couple of years later start to understand why comapanies joyfuly spent hundreds of million dollars in the early eighties to dump hierarchical databases for RDBMSes."
-----------------------------------------
My initial reaction: SHEER HORROR
My secondary reaction:  "Hey, jobs!"

:-)

John Rose
Tuesday, January 06, 2004

Actually, there is lots of interesting stuff going on nowadays
in database engines, although it is mostly "on the margins"
(as is most interesting stuff :)

Where I'm at is in really small database engines, which are
used in controllers, sensors, and other tiny devices, which
have more data management requirements than one may
typically imagine, but which are small enough that a full
blown PC environment isn't appropriate to their cost models
as they typically have only a few dollars' worth of hardware
available.  Also, traditional models of streaming sensor
data to The Big Server are falling apart under the fact that
people want more and more little gizmos everywhere, and
would prefer smarter gizmos and less single-point-of-failure
server architectures if they can be avoided.

Also, ultra-huge DBMS's are also interesting - and have lots
of the same engineering challenges as the little ones do, at
least in that huge and tiny database engines are subject to
physical constraints in ways that middle-sized database
engines aren't.  But I don't like ultra-huge DBMS work since
it tends to come with ultra-huge companies and their
associated bureaucratic ditritus...

Another development is open source.  Open source
RDBMS's are going to ultimately do the same thing to ORCL
that open-source OS's (ie, Linux) are doing to OS vendors.
Larry obviously realizes this and has been trying to
move his oil tanker of a company away from dependence
on DBMS engines, but _to_ what is still an open issue.
He tried to be Siebel, and PSFT, and SAP, and hasn't
had enormous success in any of these pushes.

foobarista
Tuesday, January 06, 2004

K - and I thought Kafka had posted to JoS. Now I'm disappointed

Nick
Tuesday, January 06, 2004

Nick,

I was deeply disappointed with Kafka. Do you like him?
I liked "Judgement" and "Metamorphosis", but everything else kind of escaped me.

K
Tuesday, January 06, 2004

We recently started upgrading all our databases from Oracle 8 to 9.2.  Why?  Stability, support, speed, scalability.  Now we churn a lot of data, we probably take in a gig of new data every day or two.  And that's  the business reason.

Now here's the reason it makes me happy as a programmer.  With Oracle 9 the engine that runs PL/SQL is finally the same engine that runs standard SQL.  That means I can put CASE statements in my select instead of having to deal with a huge number of difficult subqueries.  We actually upgraded right during the middle of the writing of a huge extract I was writing.  Suddenly my code dropped from nasty to moderately nasty and now someone else can actually make sense of it.

There will always be business reasons to upgrade, and there will always be reasons that the coders see.  Heck, the Row Level Security package is just starting to get interesting in 9 and it sure offered us some great possibilities on our database.

Lou
Tuesday, January 06, 2004

I agree that databases have become uninteresting.  But I think there is a huge untapped market out there of people who are just overwhelmed by the amount of data they're gathering.  At least thats what I'm betting my company on.

Mike
Tuesday, January 06, 2004

I think that there will be a lot of companies that do not see a reason to upgrade their database systems. Especially, if the system is not changing a lot. There is a risk when the vendor ends support but if a system has been running for several years without problems the risk can be low. 

But this will not reduce the development on Databases, it will increase the development effort. The reason is that the main revenue stream for the database vendors is upgrades, since most companies that need a database have one.  To encourage updates the vendors need to add features so they have something to shout about.  The same argument can be made about Microsoft Office but we see a new version every few years so Bill can collect the upgrade fees.

john
Wednesday, January 07, 2004

K,

I never read "Judgement", and "Metamorphosis" didn't appeal to me.  I liked "The Trial", "The Castle", and a number of his short stories, though.  Like them or not, his stories leave a lasting impression on you.

If you've ever fought an uphill battle to make changes in a bureaucratic organization, then you can relate to "The Castle".  And, if you're too lazy to read the book, they made a movie of it back in the 60's starring Maximillian Schell.  the movie version of "The Trial" starring Anthony Hopkins is also worth watching.

</end of quasi-literary aside>

Nick
Wednesday, January 07, 2004

I was also looking forward to something Kafkaesque...

However, I think you're wrong in any event, database development continues.  RDBMS isn't the only reasonable way of organising data and MDBS have shown that you can have data access methods that make sense for the application, object store, relational, network; but keep the data in a single form.

If MS and Oracle have run out of ideas, and I don't think they have, then someone else will come out with a shinier wheel which will get adopted in some form by the others.

Simon Lucy
Wednesday, January 07, 2004

*  Recent Topics

*  Fog Creek Home