>>I think I read somewhere that 99.9% of all relational database
>>installs today are less than 1T in size. As data warehousing becomes
>>more popular, this'll probably change.
I agree with you on the size. There are VERY FEW OLTP cases that run 1T of data, and it seems that most of these large ones are on Oracle, because they could do cluster support for OLTP sooner.
However, the number of data warehouses that are very large (over 1T size) is growing rather quickly. Remember that it is not raw data size, but the actual size (especially with denormalization and the number if indexes) can be several times the size of raw data. As many businesses see returns from data warehouses (their own or their competitors), there will be a lot of pressure to examine ever more data to draw more, better, or different conclusions.
However, I believe that the real growth in size will be in multi-media DBMS. The storage of pictures, sound, ... will really grow the size (but not necessarily the number of rows) greatly. IFMX already has two of the leading cases in this area (NASA EOSDIS and the Time Warner Photo Database) that have been mentioned in recent posts. I expect to see a lot more of these as people start to move from paper to on-line storage. Further, the speed of access to different data types will be critical, and I believe that the Informix architecture will offer the superior speed and flexibility for the specialty data types.
Further, the size of data means that the amount of disk storage will dwarf the CPU and memory costs for many of these sites, so the difference between NT and Unix machines may be smaller, and Microsoft would lose one of their advantages (lower cost of NT). |
