Re: Larry Ellison and million bucks!

raman batra wrote: > > Yeah, but when people use Oracle 8.i for data warehousing they would be > using aggregate tables and materialized views. If a big TB data > warehouse takes long to load, if the response time is good, who cares ? You only have 24 hrs/day right? > Our end-users want big queries (4 -5 table joins) fast. > And what happens if you don't have the pre-computed stuff in place (because it's an ad-hoc query that was not anticipated during schema creation). > I believe, Oracle is trying to take a leaf out of Informix's book and > Materialized views and pre-computed aggreggates do the same function as > summary tables which are very very useful in multi-dimensional OLAP. > That is correct. Various OLAP vendors build their own summary tables for performance reasons etc. However, the spirit of these TPC D benchmarks is a bit different IMHO. Point is: You are doing a 1 TB benchmark. You could also be doing 30GB/100GB/300GB etc. If the queries in different benchmarks are not in fact reading similar amount of data in all cases, the benchmarks will be incompatible between vendors. For example: If one would change a particular query slightly so Oracle 8i could not use the pre-computed stuff would it still run in 71 secs? The TPC-D benchmark was made to measure the performance of DW. I assume that all vendors could come up with a solution where the queries could be answered faster (by summary tables etc.) That however will not tell you how fast a database (and hardware) can actually scan/join/sort/aggregate billion rows (or whatever). Cheers, -Snorri -- Snorri Bergmann | Mail: snorri@strengur.is Strengur Inc. Armuli 7 | 108 Reykjavik Iceland | Phone: +354 550 9000 (9007 direct) http://www.strengur.is/ | Telefax: +354 550 9010