db-derby-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Michael Segel" <mse...@segel.com>
Subject RE: does Derby has some form of performance problem ?
Date Tue, 07 Nov 2006 13:35:53 GMT


> -----Original Message-----
> From: Bernt.Johnsen@Sun.COM [mailto:Bernt.Johnsen@Sun.COM]
> Sent: Sunday, November 05, 2006 3:49 PM
> To: Derby Discussion
> Subject: Re: does Derby has some form of performance problem ?
> 
> >>>>>>>>>>>> legolas wood wrote (2006-11-03 14:46:10):
> > Hi
> > Thank you for reading my post
> > i want to know whether this benchmark result is correct or not.
> 
> All benchmarks are correct. You just have to figure out what the
> benchmark actually tests and then interpret the results in the proper
> context. It may or may not be relevant for your application.
> 
[mjs] 
All benchmarks are lies unless you do them yourselves.

In the database wars of the 90's, each company would find a hardware partner
and tweak the configuration to give them the best performance. Then there
was Oracle which wrote some internals specifically to enhance their OLTP
performance.

> When it comes to H2, I would recomend that you read page 43 in the H2
> documentation about durability: "This database does not guarantee that
> all committed transactions survive a power failure".  Durability comes
> at a cost, it's as simple as that.
> 
[mjs] 
What suggests is that they consider the transaction complete when its
written to the buffer/cache prior to actual writing on the disk.

> > we are using derby as embedded database and it took a long time to start
> > the database.
> > we use a proper shutdown every time we are leaving the application.
> > does any one here used derby for real world jobs (embedded mode) that
> > database size growth to +100 MB ?
> >
> > benchmark result
> > http://www.h2database.com/html/frame.html?performance.html&main
> >
> > is it usual for derby to take 20-30 seconds to start ?
> > does startup time depend on database size ?
> 
> 
Start up time is going to be relative. And its "cheatable".




Mime
View raw message