cjackso@uswnvg.UUCP (Clay Jackson) (01/23/91)
As our Oracle applications get larger and more complex, we're seeing a number of cases where tables are getting too large to process rapidly in an OLTP environment. For example, a table that contained 10K rows fits entirely in the SGA (shared memory, for those of you not using Oracle), but at 100K rows it no longer fits. The result is that database operations (esp queries) are taking longer and longer. Our users tell us that it's "ok" to move "old" data to someplace else that would be "less" accessible (ie, they would accept slower response time and/or in some cases be willing to put up with delays while off-line media was retrieved). There are several "issues" that we're facing, and wondering if anyone else has solved some of these problems: 1) What types of reasonable (performance/cost) devices exist for large Unix platforms (we're running a Pyramid/Sequent shop) that will enable us to store and retrieve several GB of data? 2) How do we "archive" the data in such a way that we retain all of the foreign key relationships as they were at the time the data was last updated (for example, we have several tables that reference more than 3 other tables)? 3) What sort of mechanisms an we use to allow potential updates of this data? Clay Jackson US West NewVector