>I have several tables of 1 million+ records that are frequently replaced in their entirety. When this is done, they are all indexed on every single field.
>
>In one instance, indexing a 1 million record file on 183 fields can take from 5 hours (process running on server) to over 15 (over crowded LAN).
>
>I thought: Wouldn't it be cool to have a multiple key indexer that reads records only once and builds multiple keys as it goes, allowing only a single pass through the file which results in a complete CDX.
>
>If, by any stretch, someone has something like this, please let me know. If anyone else has the same need and would like to collaborate on getting it written, let me know also and we'll see if we can crack it.


I have a system that contains 3 tables frn (5,000 records) lot (500,000 records) and comp (5,000,000 records). I had the same problem with adding new records and reindexing. How I sped it up was to have all the tables brought off the network to a machine with 128MB memory and 10GB hard drive and 400MHZ processor. This machine will do all of my grunt work of clearing out all tables and adding new data and then rebuilding the indexes. The process took a long time on the network, but using this approach with all the files locally. The processes will take only one hour or less (and that includes zipping the files up for transfer back to the network where they are unzipped and put back in operation).

HTH