Archives of the TeradataForum
Message Posted: Mon, 01 Jul 2002 @ 19:24:25 GMT
| Subj: | | V2R3 - V2R4 Migration. |
| |
| From: | | Richard Perry |
We are currently working towards the migration of our Teradata Data warehouse from a 4 Node NCR 4800 (450 Mhz processors) running
V2R3.0.3 to an NCR4900 (1.4 Ghz processors) running V2R4.1.2. We have recently noticed a number of threads on this list and this has
caused us a number of concerns, mainly that once we move to the new configuration we will start to see problems with our highly
complex Business Objects queries due to issues with the optimiser in the new version. Although we do collect statitistics on the
tables in question, we understand that people have experienced issues that required further optimisation work.
Has anyone seen any specific examples of queries failing for any reason once they have migrated from V2R3.0.3 to V2R4,
specifically relating to optimiser changes. I would be interested in a feel for the type of queries that cause problems too, i.e. is
it just large multiple joins or large fact to small dimension tables, nested queries etc.
Due to aggressive timescales, we have considered upgrading our current system to V2R4.1.2 prior to going to the new box, to reduce
the number of things to resolve at changeover time. This was largely to address as many problems as we could up front, especially
optimiser issues. The big risk of course is this would be to our live system. We feel however that this would probably be a waste
of time due to the different hardware configs. i.e. we would optimise based on new software, old system and then have to do it again
when we move to the new system as the optimiser would probably do things differently again. This negates one of the main reasons for
doing it.
Any feedback on the above points would be greatly appreciated.
Richard.
| 
