Hi Chad

Chad Nantais <chad at whistler.com> wrote:
> I am building a system from the ground-up and would like to stick to 
> Squeak as much as possible to implement the system.  The system will 
> eventually generate hundreds of thousands of instances and I will need 
> quick, query-based access to these instances.  How far can I go with 
> using a Squeak image for storage before I need to use some form of 
> external storage like Relational or Object DBs?
> 
> Chad

Congratulations to you for choosing this approach!

I would recommend you to do an experiment.

Create your net of objects, fill it with random data (strings and random
numbers),
save the image, check the amount of memory it needs, restart the image
create
additional data, measure againg until you get enough information to have
an estimate.

Nowadays it is easily to work with let's say 512MB of RAM. Diego Gomez
Deck
says that he has been working with hundreds of thousands of instances.

At the moment I am using this approach but only for about 30000 objects.
I have an image size of 52MB, but because I have about 70 projects with
a lot
of graphics in it as well.  I have 192MB of RAM.

It is really exciting to work with a net of living objects you can
query. You
even can change the features of your living objects as you work (adding
instance
vars and methods). This is the *huge* difference to Java as Alejandero
Raimondo
likes to put it.



Cheers

Hannes