8.11 and unicode

[owcnc]

List,
I have completed an upgrade from XE to 8.11. During the upgrade we chose NOT to do the unicode conversion due to time constraints for go live. Has anyone completed the unicode conversion and found performance any better?? Our perfromance is fine, but we are getting a lot of web exceptions related to the JDEUNICODE.dll(Mostly coming from P42101). When the web exception occurs, it kills all the users in that callobject kernal. Right now we are averaging 4 users per kernal. We have a call open with support, but they are little help. Seems that support is getting good at telling their clients that we need to upgrade to the latest tools release or convert to unicode. I would do it if they would give me the gaurantee that our issues will be solved.

So has anyone converted to unicode and seen a huge impact on perfromance or stability???

Thanks

8.11( with out SP1) TR 8.94 I1 Intel/SQL 150 users

 

[Jogil]

We have our shared objects in Unicode and Business data and control tables in non-unicode. (8.94.N1) and not yet in production but the performance is ok at the time being.

 

[jstooge]

No ifs, ands, or buts, from a performance standpoint you will see an improvement. Stability, iffy. The reason you always will get a performance bang after the unicode conversion, is that when you DON'T convert data and control tables to unicode, the E1 middleware does a non-unicode to unicode to non-unicode conversion, on the "fly". So for any operation requiring a fetch and a write or update, you are reading in the non-unicode, converting it, BSFN activity, then a unicode to non-unicode for the update back to the DB.

The things to consider for the unicode conversion are: Any custom BSFN's will need to be visited to make them unicode compliant; timing of the conversion process (purge baby, purge) as I've seen some instances of the conversion running for over a week due to the volume of data, I always recommend doing at least one "test conversion"; and of course disk/dasd/storage requirements...depending on your RDBMS system, the increase in space can go from almost nil to 40-60% more space required.

Getting back to performance, in testing done by JDE/PSFT, by converting to unicode, you should see anywhere from a 25-45% bump in performance. Last client I saw go through this, took a problematic UBE that was running for over 90 minutes, down to less than a minute (though it was custom, so only the supreme being or the spagetti programmer knows what it was doing).