VIRTUAL INDEXING SCHEME??

Does anyone know of a 'virtual indexing' scheme provided by (or available for purchase with) any RDBMS packages? Let me explain further: Our group manages water resources data, which is comprised largely of time series data. Each record in a time series table will consist of: Site ID (e.g. integer identifier for a dam) Date (date the value was recorded) Value You can imagine that LOTS of space is taken up storing dates which are really implicit in the time series itself. That is, the values are recorded at regular intervals, so the date can be ascertained from a few pieces of information, such as start date, increment, etc. The virtual indexing scheme I'm referring to (and have heard rumor that it exists) would be comprised of a data file containing 1) a matrix of values (sites and values); and 2) a set of meta information which describes the time increments (or whatever the indexing scheme happens to be) of the data. The file would be compatible w/ SQL (or some DML interpreter). Hence, a SQL end-user could query the data, asking for all values for some given site between some range of dates. The SQL interpreter could go out to the data file, and, given the meta information in the data file, query its contents appropriately. So, the file would *appear* as a regular data table to the end user, but its storage would be much more compact as the time series would be described in a few lines of meta data, not as a column in each row of data. Does such a package exist?? Or something similar which could provide more efficient management of the types of data I have mentioned?? Please respond by e-mail to: marra@cadswes.colorado.edu Thanks, carol