Have you looked at IFMX, especially what used to be called the XPS version 8.2 (they renamed the entries in their line recently). Consider the following:
>>3) Good performance in multi table joins, even when joining 14
>>tables together. This was important for us because Marketing
>>frequently requires 12 months of data to even begin working on a
>>decent model. In Oracle we have to avoid this by spending the time
>>and space to pull all the desired months into a single record for
>>each person or product.
In particular, IFMX allows you to build multi-table join indices to reduce part or most of the time involved in complex joins, and you do not have to do the Oracle thing of "denormalizing" the data for performance.
>>3) Totally hit a wall at 600Gb (though now this may no longer be
>>true)
What is the 600G limit based on? cleaned, raw data? sum of loaded data? with/without indices? IFMX has reference sites with more than 1T of raw data (the smallest size used to characterize data sizes). Cornell University is doing research on a 500 node IBM SP2 that will go well beyond that size, to see what kind of things warehouses can do. Incidentally, one of the key size characteristics of IFMX is the limit of about 60T as the maximum size of a single table.
>>4) Loader much slower than Oracle's (Again, may no longer be true)
IFMX loaded seems to be 50% faster than Oracle (especially Oracle 7) based on same or similar equipment benchmarks (see the TPC/D benchmarks).
>>5) Forced us to use the Star schema, which was counter to our
>>experience and infrastructure.
I do not think IFMX would force this kind of requirements on you.
I list the above not necessarily to convert you, although I still have IFMX stock and have a soft spot in my heart for them (I used to work there). The above also would demonstrate the problems that Red Brick has in getting into larger sites, or into sites that expect to scale to larger sizes (many of the companies expect their warehouses to grow very significantly over time). In other words, Red Brick must get past IFMX and IBM just to get to 2nd place on the list (assuming ORCL is first), and many companies do not want to do extensive evaluation past 2 finalists.
Mark
PS. To those who look at competitors, MSFT is in at least 5th place technically for complex query and other warehouse support (behind ORCL, IFMX, IBM, and SYBS), and simply does not have much in terms of references because they are only now adding the kind of technical features needed to support warehousing (and who knows when they will actually ship?). |
