It's really useful for archiving if the key contains a date -  build the
algorithm on the date in the key and have a separate partfile for each year
(or quarter or month - you get the idea),

When you want to archive old data you just drop the partfile concerned -
remembering to add new partfiles for new dates as you go forward (this bit
can be automated),

ALWAYS remember to have an "ELSE" or "CASE 1" on your algorithm just in
case, and monitor the specified partfile to fix whatever problem caused data
to be placed in there (incorrectly - or maybe you just forgot to create a
new partfile soon enough).

Caution - I have seen some grief where partfiles are added that already
contain data - and data that does not hash to this partfile under the
distribution algorithm rules. OK - it doesn't happen often - but it can
cause grief and confusion - just be careful.

Which algorithm to use ? sorry - answer is "it depends"..... 

1/ do you want even distribution - ?
2/ do you want context distribution (such as a date or a zip code)?
3/ How unique is your key?
4/ what are your data volumes?

It all depends.........

Regards

JayJay
-------
u2-users mailing list
[email protected]
To unsubscribe please visit http://listserver.u2ug.org/

Reply via email to