roy@arc.ab.ca (Roy Masrani) (11/22/90)
We have a PC application that requires a high performance PC shell for under $5K. Recently, we made a decision against PC-KAPPA due to problems we were having with performance (even though it meets our needs w.r.t. functionality). Basically, we are looking at a minimum performance expectation of around 100 - 200 rules per second based on crude estimates of the application. We are considering CLIPS but have inconsistent data ranging from 50 rps to 600 rps. Does anyone have performance data on a 486/25 based machine? We are also concerned with CLIPS' rule language (i.e. not easy for a non-programmer to modify/maintain). Any feedback on other shells out there would be greatly appreciated. I imagine this is not a new question, so I would appreciate getting a synopsis of previous discussions. Thanks in advance. Roy Masrani...3rd Floor, 8street NE...Calgary,AB,Canada...403-297-2656
mikeo@sae.sae.com (Michael Ovington) (11/29/90)
>We have a PC application that requires a high performance PC shell >for under $5K. Recently, we made a decision against PC-KAPPA due to >problems we were having with performance (even though it meets our >needs w.r.t. functionality). Basically, we are looking at a minimum >performance expectation of around 100 - 200 rules per second based on >crude estimates of the application. > >Any feedback on other shells out there would be greatly appreciated. >I imagine this is not a new question, so I would appreciate getting a >synopsis of previous discussions. > >Thanks in advance. > >Roy Masrani...3rd Floor, 8street NE...Calgary,AB,Canada...403-297-2656 Software Architecture and Engineering (the company I work for) has an expert system shell called KES. It runs on almost anything, including the PC under DOS and OS/2. On a 33Mhz Northgate with 32k cache under DOS, KES fires more than 1000 rules/sec, where the rules in the test case are simple, e.g. if a = b then c = d. We don't have a 486 in house so I don't have any timings on one. I presume it would be even faster. Of course, more complex rules will affect affect the results. I know that we can send out evaluation copies to prospective customers. If you would like more information, please contact Ricki Kleist for marketing info (price, training, evaluation copies etc.) David Kiker for technical info (features etc.) both can be reached at (703) 276-7910 Michael S. Ovington Software A&E (703) 276-7910 1600 Wilson Blvd, Suite 500 uunet!sae!mikeo Arlington, VA 22209 mikeo@sae.com