db-derby-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Jianhui Jin" <ced1...@leeds.ac.uk>
Subject RE: Derby vs. HSQL
Date Thu, 09 Sep 2004 15:18:36 GMT
Hi all

I am newbie to sql as well as jdbc (did a bit coding before). Also
interested in the performance of retrieve data or update data.

I run the same test with slightly changing in the code.
Couldn't bear with 10,000 round, so set 1,000. it seems hsql finish it
in a flash, whereas cloudscape need 172 seconds. 

I guess, hsql use as much memory as possible whereas cloudscape does
everything in the disk. Otherwise, no code could generate so much
difference. But the test code only try to use small amount of data, 3
columns with 4 * 1000 or 4 * 10000 row, if someone try to increase the
data in one table, any changes? I guess HSQL have a very good engine to
load data from disk to memory, vice-verse and use memory to keep as much
as data as possible.


Another question is strangely I couldn't find the created database file
or folder in local driver "D:/temp/testdb", whereas hsql put my database
in? any suggestions.


-----Original Message-----
From: Steen Jansdal [mailto:steen@jansdal.dk] 
Sent: 09 September 2004 12:21
To: Derby Discussion
Subject: Re: Derby vs. HSQL


Wes Johnson wrote:
> Greetings,
> 
> I'm in the process of evaluating databases for use in an application
> that I'll be writing in the near future.  As performance is very 
> critical, I'm only considering in-memory and in-process DBs.
> 
> For starters, I decided to pit Derby (actually Cloudscape 10.0) 
> against
> HSQL by running some extremely informal benchmarks.  Basically, what I

> did was in a loop of 10,000, insert 4 rows in a table and then select 
> all rows.  In other words, run 40,000 inserts and 10,000 selects.
> 
> The results greatly surprised me; for the in-process mode, HSQL came 
> out
> to be 3+ times faster than Cloudscape!  I attempted to tune the latter

> by increasing page size, etc., but I wasn't able to improve 
> performance.  I did notice that HSQL was utilizing almost all of the 
> CPU, while Cloudscape was using hardly any (with java.exe hovering 
> around 3% of utilization).
> 
> I would tremendously appreciate any suggestions for tuning Cloudscape
> such that I can achieve the *correct* performance. :)  Or, if anyone
has 
> compared Cloudscape to HSQL already (a web search turned up very 
> little), I would be very curious to hear about it.
> 
> Thanks in advance!
> 

Could you post the source for your benchmark program?

Steen

Mime
View raw message