Hello Jess-users,
On recommendation of Mr. Friedman-Hill I am posting my question here...
I have attached the original email...
thanks,
daniel


I think Daniel Theobald wrote:
> 
> Hello Mr. Friedman-Hill,
> 
> My colleagues and I are currently exploring the use of Jess in the area
> of infection control surveillance, and I was hoping that I could trouble
> you with a question about it....

No problem.

> 
> For infection control, we must be able to review large volumes of data,
> conceivably millions of records stored in a database.  Also the number
> of rules will be quite large (The system will auto generate many
> rules).  Do you know of any projects that have used it on this scale?
> (How do you deal with not having all the records in memory at a time?)

This is something that you'd have to do explicitly, loading and
unloading subsets of the data. If the data is cleanly separable into
"modules" then of course this is easy. If it's not, then there's a problem.


> Would it be reasonable to use it on this level, or would it make more
> sense to build the rules as SQL queries from a performance point of view
> (which could take advantage of the indexing etc. of the database, etc.)

The strength of a relational database would (surprisingly, perhaps)
not be that it would be faster, since the Rete algorithm builds more
and more complex indexes than any database does. This comes at a cost,
of course: first, that all the data being reasoned about is in-core at
once, and second, that it takes more memory to store the data than
you'd conservatively estimate.

Millions of facts would lead to a heap size of (minimally) hundreds of
megabytes, so that your problem would not run on a Palm Pilot. But
gigabyte memories are becoming almost commonplace, so, depending on
your target environment, this still might be feasible.

I know Jess has been used for some fairly large problems, but probably
none quite so big as yours. You might solicit advice from the
jess-users mailing list ([EMAIL PROTECTED]) to see if you can get
any more data points.

> 
> Thanks for your help!
> Daniel A. Theobald
> 
> 
> 
> 
> 


---------------------------------------------------------
Ernest Friedman-Hill  
Distributed Systems Research        Phone: (925) 294-2154
Sandia National Labs                FAX:   (925) 294-2234
Org. 8920, MS 9012                  [EMAIL PROTECTED]
PO Box 969                  http://herzberg.ca.sandia.gov
Livermore, CA 94550


Reply via email to