Subject: Keys, Large Transaction Files,
Just recently ran into an interesting phenomenon - I was working with a file
with compound keys and the selects over a date range were atrocious. I copy
the data to a new file, using sequential keys and the selects averages
200-2000X faster (for the doubters, I have to say is the actual # were
something like 197X to 2070x, the second # being as second select after the
data was cached). The avg length of key on the file was 32 characters. The
avg length of a sequential keys was about 5 characters. The fields was a
'date' field. The field was indexed. The range of the select was 2 days.
It seems there's a Unidata threshold large key sizes exceed with indexing
that kills peformance.
Also, sequential keys hash the best. I managed a file with 80M records at
another site and had no problems with file sizing or overflow.
Brad
_______________________________________________
U2-Users mailing list
[email protected]
http://listserver.u2ug.org/mailman/listinfo/u2-users
