[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Object creation is slow. Disk Access?

On Sun, 20 Feb 2000, William Uther wrote:
> Hi,
>   I finally got time to finish debugging my ozone app.  It was an annoying,
> intermittant bug based on using == to compare proxies instead of isSame().
> sigh.  Anyway, I'm now to the point of optimizing for reasonable speed.
> 
>   My app uses Ozone objects internally because I want to use it in
> situations where I may end up with a large object graph and I don't want it
> all sitting in memory.  Unfortunately that means it creates and destroys a
> lot of Ozone objects in a single transaction (10,000 objects for a small
> transaction (I did say a lot :)).
> 
>   While accessing Ozone objects is fast, either creation or destruction of
> Ozone objects (not sure which) seems to require a disk access.  This is
> making my app slow to a crawl.
> 
>   Is there any way to tell ozone to only write to disk on commit?
WizardStore and ClassicStore have caches that hold "active" clusters/objects.
They use some kind of LRU to decide which objects are "active". If your working
set of date (objects) fit in the cache, ozone *does* only write on commit.
But the size of WizardsStores cache for the 0.4dev3 release defaults to 2MB
which is probably to low to hold all of your 10k objects. You may change
this by editing the config.properties file of the database or use default
properties when installing the database using ozoneInst. If you increase cache
size you also have to to increase Java heap of the server. Currently it is set
to 32MB (see bin/ozone). I don't know a good size for say 5MB ozone object
cache. Test it out and let as know.

> 
> later,
> 
> \x/ill            :-}
> 
> P.S.  My app is fairly small, but exercises the ODBMS heavily.  As soon as
> I get it cleaned up I'll release it Open Source so you can optimize against
> it if you wish.
Great, a new ozonie... :)

> 
> P.P.S.  Each time I've posted on this list it's been to point out something
> is broken.  I should balance things a little by congratulating Falko et al.
> on their great work and thanking SMB for making the project open source :).
:) The success of the entire Ozone Database Project thing highly, highly, highly
depends on the level the users and developers try to get envolved, try to
understand ideas behind, try to setup project on top of ozone, and of course
sharing their experiences. I appreciate all posts even if the tell me that
something is broken ;) So please, keep it coming! Thanks Will.


Falko
-- 
______________________________________________________________________
Falko Braeutigam                         mailto:falko@softwarebuero.de
softwarebuero m&b (SMB)                    http://www.softwarebuero.de