home bbs files messages ]

Forums before death by AOL, social media and spammers... "We can't have nice things"

   comp.ai      Awaiting the gospel from Sarah Connor      1,954 messages   

[   << oldest   |   < older   |   list   |   newer >   |   newest >>   ]

   Message 496 of 1,954   
   Babis to Elpico   
   Re: Rules engine performance is my bigge   
   22 Nov 04 16:45:04   
   
   From: marmanis@computer.org   
      
   Elpico wrote:   
   > Hi all,   
   >   
   > I'm investigating whether rules engines would solve a problem within   
   > my company. We're a large financial institute, and the existing rules   
   > are hard-coded bits of C and assembler (nice) compiled and run on a   
   > IBM mainframe.   
   >   
   > Biggest problem for us obviously is the time taken to   
   > update/change/modify rules - currently weeks/months. Now, we are   
   > currently looking to migrate a large chunk of or core processing   
   > systems, but without affecting the performance (peak processing of   
   > approx 7000 transactions per second, with an estimated rule base of   
   > 1500 - we're still investigating the complexity of the rules (depth,   
   > rule-fire ratio, etc).   
   >   
   > We're looking at existing systems vs developing our own solution, and   
   > was wondering if anyone had any suggestion on products that could   
   > perform.   
   >   
   > Given that the data is constantly changing and the partial rule   
   > matching will be low, will the rete alogorithm  (for example) be   
   > redundant in our case? Any suggestions on alternative   
   > approaches/products?   
   >   
   > Cheers,   
   > Elpico.   
   >   
      
   Hi.   
      
   Look at Drools. It is LGPL.   
      
   You may be able to use only what you need and get 5000 trx/sec or so.   
   It depends on many factors, of course. I am making several assumptions   
   to get to that number.   
      
   I used a custom implementation of RETE 4 years ago for a trading   
   platform.   
   It was matching bids/asks that were attribute based. We got 15,000   
   trx/sec   
   matching rate in Java.   
      
   The rete algorithm and its variants are the fastest known algorithms   
   for rules engines; well, for those that nit-pick, they are the fastest   
   publicly known algorithms.   
      
   If you do not have many rules (1500 is not much, it's quite a small   
   number actually), you do not change them frequently (every second or   
   minute), and you want to get the highest performance then RETE is   
   probably your best choice.   
      
   Best,   
   Babis   
      
   [ comp.ai is moderated.  To submit, just post and be patient, or if ]   
   [ that fails mail your article to , and ]   
   [ ask your news administrator to fix the problems with your system. ]   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)   

[   << oldest   |   < older   |   list   |   newer >   |   newest >>   ]


(c) 1994,  bbs@darkrealms.ca