> Lionel van den Berg, 03.03.2008 11:19:
>> I'm looking into Java DB, and although Sun claims that it can be a
>> full- scale DBMS I'm wondering what the views on this are? Can Java DB
>> be used for large databases?
>
> Define "large"
Sigh, I knew that was coming. I'm not sure, I'm not a database person, I
do enough to get by.
Let's say 10 tables with an average of 4 fields with each table having up
to 10 million entries.
Not sure if that constitutes large, I'm sure it's not to google.
Thomas Kellerer - 03 Mar 2008 12:16 GMT
Lionel van den Berg, 03.03.2008 12:47:
>>> I'm looking into Java DB, and although Sun claims that it can be a
>>> full- scale DBMS I'm wondering what the views on this are? Can Java DB
[quoted text clipped - 8 lines]
>
> Not sure if that constitutes large, I'm sure it's not to google.
OK, I would think that qualifies as "large" (at least not as small :) )
I have no experience with Derby using such a setup. I would be a bit hesitating about using it for such a database. But I guess this can only be answered by setting up a testdatabase of that size.
You might want to post that question to the Derby mailing list, support over there is quite good.
Thomas
Lionel van den Berg - 04 Mar 2008 13:55 GMT
> You might want to post that question to the Derby mailing list, support
> over there is quite good.
Thanks. Sort of trying to get the unbiased opinion, but it's still worth
a try :).
Arne Vajhøj - 03 Mar 2008 21:46 GMT
>> Lionel van den Berg, 03.03.2008 11:19:
>>> I'm looking into Java DB, and although Sun claims that it can be a
[quoted text clipped - 7 lines]
> Let's say 10 tables with an average of 4 fields with each table having up
> to 10 million entries.
If it is integer fields then 10*4*10M = 400 MB which is a small
database.
Even though I have not tested Derby with a 400 MB database, then I
am sure it can handle it.
Arne
Arne Vajhøj - 03 Mar 2008 21:46 GMT
>>> Lionel van den Berg, 03.03.2008 11:19:
>>>> I'm looking into Java DB, and although Sun claims that it can be a
[quoted text clipped - 13 lines]
> Even though I have not tested Derby with a 400 MB database, then I
> am sure it can handle it.
Translation:
MB = small
GB = medium
TB = large
PB = arghhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhh
Arne
Daniel Pitts - 03 Mar 2008 22:30 GMT
>> Even though I have not tested Derby with a 400 MB database, then I
>> am sure it can handle it.
[quoted text clipped - 5 lines]
> TB = large
> PB = arghhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhh
You mean: PB = Google

Signature
Daniel Pitts' Tech Blog: <http://virtualinfinity.net/wordpress/>
Christian - 04 Mar 2008 00:49 GMT
Daniel Pitts schrieb:
>>> Even though I have not tested Derby with a 400 MB database, then I
>>> am sure it can handle it.
[quoted text clipped - 6 lines]
>> PB = arghhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhh
> You mean: PB = Google
nope not Really.. I googled a and found some blog entry from 2006
estimating Googles data in compressed form to 220TB. I doubt it grew to
a PB since then.
http://googlesystem.blogspot.com/2006/09/how-much-data-does-google-store.html
Daniel Pitts - 04 Mar 2008 02:14 GMT
> Daniel Pitts schrieb:
>>>> Even though I have not tested Derby with a 400 MB database, then I
[quoted text clipped - 13 lines]
>
> http://googlesystem.blogspot.com/2006/09/how-much-data-does-google-store.html
Compressed form...

Signature
Daniel Pitts' Tech Blog: <http://virtualinfinity.net/wordpress/>
Arne Vajhøj - 04 Mar 2008 02:28 GMT
>> Daniel Pitts schrieb:
>>>>> Even though I have not tested Derby with a 400 MB database, then I
[quoted text clipped - 15 lines]
>>
> Compressed form...
I am a bit surprised that it is no more data.
According to:
http://research.google.com/archive/disk_failures.pdf
Google has >100000 disks 80-400 GB.
100000*100 GB is 10 PB.
Arne
Thomas Kellerer - 03 Mar 2008 22:45 GMT
Arne Vajhøj wrote on 03.03.2008 22:46:
>>> Lionel van den Berg, 03.03.2008 11:19:
>>>> I'm looking into Java DB, and although Sun claims that it can be a
[quoted text clipped - 10 lines]
> If it is integer fields then 10*4*10M = 400 MB which is a small
> database.
The actual data size is not necessarily an indicator whether a database is
"large". 10 Million rows (per table) could be a problem, if e.g. the access path
to retrieve a subset is not well chosen by the engine and it has to scan all
rows in order to retrieve e.g. 5 of them.
If the number of updates and parallel reads is very high, but the design does
not allow for concurrent updates and reads ("writers blocking readers") then the
performance of the application will suffer as.
Btw: writers blocking readers is a problem even for some of the "big" names
Thomas
Arne Vajhøj - 17 Mar 2008 03:52 GMT
> Arne Vajhøj wrote on 03.03.2008 22:46:
>>>> Lionel van den Berg, 03.03.2008 11:19:
[quoted text clipped - 16 lines]
> access path to retrieve a subset is not well chosen by the engine and it
> has to scan all rows in order to retrieve e.g. 5 of them.
I would phrase it as: large/small depends on size, but whether
a database is suited for a given problem may depend on many other
factors than the size of the database.
Arne
francois.orsini@gmail.com - 04 Mar 2008 21:55 GMT
Hi Lionel,
Java DB (Apache Derby) can definitely handle such a dataset size. We
have run some performance tests in multi-user mode with databases >
30+GB.
You can also check about feedback on using large datasets at:
http://www.ohloh.net/projects/3816
Derby is suitable for medium to high-level departmental type of
databases. It was built originally (back at Cloudscape 10+ years ago)
with no particular limitation in mind. The engine has always been
multi-threaded, with a cost-based optimizer and a storage engine that
can handle large datasets. Of course and as some people have pointed
here, large datasets can mean different things to a lot of people.
If you still need to, I would recommend you to post on the derby-user
mailing list, which is very active by the way:
http://db.apache.org/derby/derby_comm.html
You can also browse the list via nabble at
http://www.nabble.com/Derby-f356.html
Regards,
Francois Orsini
> > Lionel van den Berg, 03.03.2008 11:19:
> >> I'm looking into Java DB, and although Sun claims that it can be a
[quoted text clipped - 10 lines]
>
> Not sure if that constitutes large, I'm sure it's not to google.