Hmmm...I am a bit familiar with DASD and I know where you're coming from...but I really don't think that NTFS works that way...if an NT expert wants to validate the following, GREAT!!

Here goes: Everytime you access a piece of data, NT is going to try to buffer the next 'n' blocks of data in an either logical or sequential manner. Therefore, if the database is physically fragmented, one of two things would happen -- either it can't buffer because the next physical block of data is from some other file and performance is degraded -OR- disk thrash becuase it has to go looking for the next sequential block to buffer and performance is degraded....see what I mean?


>Well, John, I *hope* that I mean more than well (readily reminding you that I do not (repeat: not) know NT at all). . .
>
>But I *do* know mainframe DASD quite well - the structure of the VTOC and the allocation/access of "extents" (which is what fragments of any file are called there).
>Now I can't imagine the VTOC being any more efficient than NT's way of keeping track of storage, and from another comment offered on this thread I might surmise that it might be similar.
>And on a mainframe system, notorious as a REAL multi-tasking environment for at least the 30 years that I have worked on them, *DO* benefit from fragmented files. In fact it can make a huge difference. And there *are* (indirect) ways to control that fragmentation.
>
>I readily acknowledge that fragmentation is BAD in a UNI-tasking environment, but I continue to question that "fact" for multi-tasking PC environments.
>
>Cheers,
>Jim N
>