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

Re: Anyone here

Sorry, sorry, sorry for the slow response. I was busy preparing our
booth at CeBIT in Hannover.

Tim Brown wrote:
> 
> Hi,
> 
> I haven't seen a message from this list in a long time.  is
> it still active?
Yes.

> Is anyone else seeing what I am?  I have done extensive
> testing of ozone and have noticed that if I attempt to write
> large objects to the database it takes forever.  My project
> is a space game.  So I write the following:
> 
> starDataImpl - Has about 10 fields on of which is a Vector
> holding no more than 10 small objects.
> 
> planetDataImpl - each planet in it has lots of data.  This
> object holds all of the sector objects for each planet and
> can have thousands of objects.  Creating the planetData
> object in ozone is fast but adding planets takes forever.
> Creating a 4 star universe takes 5-10 minutes.  I need to
> create 100-200 stars really.
> 
> The dataImpl objects are modeled somewhat after the Garage
> example except that the individual objects are not
> ozoneRemote objects but rather serializable objects.
> starData holds stars and planetData holds planets.
> 
> Has anyone succeded in managinf large objects in ozone at a
> reasonable speed?  At this point I am thinking I need to use
> something else.
> 
> If anyone is interested in seeing the code, it is available
> at my URL below, just click toward new game and you can
> download it.
Unfortunatelly I am unable to start something. What I have to do?

Anyway, serialization is the performance bottle neck in ozone. On my
alpha box ozone needs 32 seconds to build 1000 OzoneObjects where each
object contains 100 regulare Strings. It's just a benchmark and the
methods are very simple so ozone needs much of the time only for the
serialization. The size of this database is 2.4 MB.

But 10 minutes are to much even if a planet contains thousands of
serializable objects. Did you adapt the size of the cluster space and
the size of the clusters for your needs? In any case the cluster space
should be able to hold the working set of a client. In your case a
planet seems to be the working set of a client.

Please tell me what I have to do to test your code and I will get back
with more information.


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