As far as I know, the current forward chaining implementation needs to
keep all facts into the RAM memory, so you can't hold 1 TB facts in
2-4GB RAM memory...

You could try streaming it into some sort of rules-based filter
configuration to avoid getting to much into the working memory and then
apply a second rule set on all the facts that made it through the
original filter.

They are currently implementing backwards-chaining, maybe that can could
also be integrated with a database that holds the terrabytes?

[EMAIL PROTECTED] wrote, On 2006-11-10 9:37 AM:
> What are you trying to achieve when you say your task is 'log processing'
>  
> Will ALL Log entries(objects) be of interest
> Assuming NOT - you can (maybe) write some Rules which filter your input 
> so that ONLY those Log entries you need to process are committed to the 
> working memory
> (you will probably have to Load each Log entry into the working memory & 
> have your 'Filter Rules' Retract Log entries that you are not interested in)
>  
> Subsequent Rules can process the 'Real' Log entries however you need to
>  
> Haven't tried any of this myself  - BUT I think you should be able to 
> make it work
>  
> 
>      
>      
>      
>     ------------------------------------------------------------------------
>     *From:* 张茂森 [mailto:[EMAIL PROTECTED]
>     *Sent:* Friday, November 10, 2006 2:46 AM
>     *To:* [email protected]
>     *Subject:* [drools-user] Some thoughts about drools
> 
>     Hi All:
> 
>     After I have read the drools user guide, and now my work is about
>     log processing. I want to know whether rule engine is proper to be
>     used to do log processing.
> 
>     1         Log can be a very large “fact” set (even tera-bytes is
>     possible). Could Rule Engine achieve the requirement of performance?
> 
>     2         As far as I know, I must wrap each log into object then
>     send it to workmemory of Rule Engine, Does it need? Could I have
>     some more lightweight solution?
> 
>     Any suggestion would be appreciated
> 
>     Thank you!
> 
>      
> 
>      
> 

-- 
With kind regards,
Geoffrey De Smet


---------------------------------------------------------------------
To unsubscribe from this list please visit:

    http://xircles.codehaus.org/manage_email

Reply via email to