This may help (or perhaps not).... In teaching Unidata courses I have heard several comments over the last year or so about large dynamic files that have apparently stopped working correctly. The two that I have been able to examine in any detail both showed that the system appeared to be splitting groups when it did not need to, resulting in vast amounts of dead space and an average load a long way below the merge load value. Unfortunately, I do not still have the analysis as this was left with the customer to forward to IBM for further investigation.
It looks to me as though there is a bug there somewhere but, because IBM do not publish any detailed internals information for Unidata, it might just be a problem with the underlying principles. In general, a dynamic file will never be as good as a perfectly tuned static file but it should be better than a static file that has been allowed to degrade through lack of retuning. Personally, I am a great believer in dynamic files and rarely use static files when working with UniVerse. We have a UniVerse client with an application that has over 100,000 dynamic files (honest!) and has never experienced a problem. Ok, most of these files are small and only a few are in use at any time. Because of the above mentioned issues, I am more wary of dynamic files in Unidata. Martin Phillips Ladybridge Systems 17b Coldstream Lane, Hardingstone, Northampton NN4 6DB +44-(0)1604-709200 ------- u2-users mailing list [email protected] To unsubscribe please visit http://listserver.u2ug.org/
