Re: HELP! SE versus Online AND is ONLINE faster than SE?

Ed Puska asked about the advisability of changing to ONLINE. I understand, there are two primary reasons given for changing to ONLINE. a) To provide an opportunity to improve data integrity (if your application does implement commit/rollback, which is not always the case) but at the cost of significantly higher system admimistration cost. b) Improve throughput It is this latter I wish to discuss. THe FAQ and various discussions in this new group claim a 2 to 3 times factor in improvement (I assume this means that given there is no CPU bottleneck, then if a report takes 60 minutes to run with SE, it will run in 20 minutes with ONLINE, if the response time is 10 seconds for an enquiry using SE, it becomes 3 seconds using ONLINE). It is these claims I question. I have NEVER seen any hard evidence that moving to ONLINE improves throughput. The reports that I have seen in this newsgroup only achieved performance improvements from ONLINE after adding significant resources to their system, primarily 3 or more disks. One then has to wonder if they would have seen similar improvement from SE, after investing in additional resources. I know of a case where a change was made to ONLINE at the cost of extreme aggravation during the changeover phase. Two disks were added (just to be sure there would be an improvement in throughput) to make a complement of five for a 150Mb database. Database tuning was performed by Informix staff (just to be sure there would be an improvement in throughput). The result of all that - there was NO USER PERCEPTION OF AN IMPROVEMENT IN THROUGHPUT. I do not consider the above experience definitive, as there was no rigorous before and after testing of performance or operating conditions. Nevertheless, in order to be convinced that ONLINE does provide a significant performance improvement requires a well documented before and after test case involving a database of at least 100Mb. Perhaps a TPC benchmark comparison of SE and ONLINE. We can be quite confident that Informix has conducted such tests. We can be equally confident that if the results are not available, there is no performance improvement to be gained by investing in ONLINE rather than SE. We might suspect the results are not available as such results have not been presented or referenced by any of the Informix `newsgroup watchers' when performance issues are periodically discussed here. I would be the first to applaud if the above conjecture can be shown to be incorrect. Regards to all Li Chuan -- Li-Chuan Dunlop -- s925095@minyos.xx.rmit.edu.au