Hi Has anyone got experience with UniVerse distributed files WRT performance? That is, any idea whether a file distributed across 10 parts gives slower performance than one distributed across only 2 parts?
(I ask because I am comparing both scenarios and, so far, the 2-part file performs better (running a program that checks each record, but also when doing a COUNT) ...or might it just depend on system busyness at the time? How the overflow occurs across 10 parts might also influence the results(?). It's a test box so doesn't have a large number of users - around ten (but it depends what they're doing, of course). The 10-part version of the file takes >5 hours to be checked by our check program, while the 2-part takes about 2.5 hours (although I'm running another timing test now, to confirm). A COUNT takes 26 mins on the 10-part, but only 22 mins on the 2-part. Each distributed file is 7GB, with over 3 million records (same records for both files). The file type and separation are the same, while the modulo is relative to the number of parts.) Thanks, LeRoi LeRoi Keiller Technical Support Ultradata - Vision to Reality +61 3 9291 1700 www.ultradata.com.au ------- u2-users mailing list [EMAIL PROTECTED] To unsubscribe please visit http://listserver.u2ug.org/
