Joerg Plewe's Blog

March 2005 Archives

A new DB benchmark in town

Since I've been investigating on persistence recently, I found that performance might be an issue, esp. when thinking about embedded databases.

So I organized a race (very obviously inspired by db4o's Formula1 tutorial) and put the results online in form of the SourceForge project PolePosition

Currently I lined up these database systems for some training laps:

• jdbc
  • MySQL
  • Apache Derby (Cloudscape)
  • HSQLDB
  • McKoi
• db4o
• Hibernate

I started from scratch without having much experience in any of the databases tested. So the code is pretty much unbiased and shows what an average programmer might achieve performance-wise. On the long hand, each database should use the best code possible. For I will never be an expert for that variety of database systems at the same time, making the project opensource gives the possibility for the dedicated experts out there to optimize for their resp. favorite. So please let me know where the code is poor and where improvements are mandatory.

The preliminary results are quite mixed without having one single winner for all purposes. Just take a look at the results yourself.

There are other important benchmarks which are not part of the results published. The 'hassle' or the 'learning curve', the 'lines of code' or 'time for debug' benchmark e.g. just to mention a few. As mentioned earlier, db4o is hard to beat there....

There are more things to come. Anybody volunteering for the JDO team?

Everybody have fun with the code and let me know what you think.

Will federate databases die out?

These days I'm fooling around with some database technologies.

Just one thing came to my mind when I was observing my own daily work. Do I query federate databases often? No, I don't. Instead I use google. That's my standard query tool and it is really powerful. Everybody does. Nobody can live without any more.

Today, federate databases are the backbone of most applications. They maintain prestructured data to answer prestuctured queries leading to prestructured results in fixed-form user frontends (well, roughly :) ).

Maybe even more valuable data (or should I say knowlegde) floats around in zillions of Word documents, presentations, emails. Nobody reads them. Some smart companies already started to install internal search engines maybe complemented by taxonomy/classification engines in even smarter companies. Will they be able to replace federate databases one day?

On starship Enterprise, the question 'What do we know about planet Omega-Haribo7?' can be answered giving exactly the information the questioner wanted. Lifeforms, atmosphere, latest wars. I'm pretty sure it's not just some voice recognition combined with a MySQL server to get that done.

Will the future decide the battle between OODBMS and RDBMS against both? Will unstructured information be the key for future enterprise databases? Will GUI-forms vanish? Will this - in the end - lead to completely new paradigms in programming languages, databases, UI ...? Just a thought before weekend.

This work is licensed under a Creative Commons License.