Hi George,

>>However, if you look at specific cases where 90 % of the disk I/O has to do with data retrieval of logical grouped record, suchs as 'dossiers' (e.g. patient records in health information systems, insurance dossiers etc), the fragmentation of the logical 'dossiers' theoretically is far worse than fragmentation of the physical tables.

>>You've got to damn well identify such case, before you draw your conclusions in order to fragment or not to fragment tables in your databases.

>See MESSAGE#737831

>I've drawn my conclusion based on solid computer science principals. I've based them on my knowledge of both the hardware and the various possibilities regarding the design and implementation of the file system. I'm satisfied that they are correct.

I'm not sure what you're trying to say here with regards to the statement I've made. I do agree that in most circumstances unfragmented files do have a (theorectical) advantage. However there might be spcific situations where fragmented files (and unfragented logical dossiers) are better performance wise.

I'm not sure what you're trying to say with: I've drawn my conclusion based... . Well I do know I have a degree in CS and my conclusions are both drawn on my knowledge of the storage system (both NTFS and FAT) as real life experience with systems where logical dossiers are stored.

I do think our conclusions do not differ too much, but we do have a different opinion of the circumstances where fragmented files (unfragmented dossiers) are better performance wise.

But again this is an academical question because:
1. We don't have much control how files should be fragemented
2. With HD and OS chaching it is difficult to measure these things.
3. With RAID controlers or tables stored on differen HDs, seek time schemes get more complicated to determine.

Walter,