RE: Delete Data older than exactly 2 years
- From: "Mark W. Farnham" <mwf@xxxxxxxx>
- To: <rjamya@xxxxxxxxx>, <smishra_97@xxxxxxxxx>
- Date: Fri, 20 Apr 2007 10:18:30 -0400
I think a lot of industries work off comparitive week analysis, and that no
particular calendar time grouping is *always* better than another.
Cohesion of weeks may even be more important than cohesion of years to the
extent that in some industries years are defined as 52 weeks most years
except when the mismatch between 365.25 and 52*7=364 adds up to needing an
extra week about every 5.6 years on average. In fact weeks are consistently
of the same duration if you allow exception for leap seconds and the like,
while months vary wildly in duration from 28 days to 31 days (albeit in a
predictable pattern).
Matching your archiving time frame to your delete (or partition unplugging)
strategy and basing that strategy on the natural or accounting rhythms of
the data collection system is often a good choice for information life cycle
management. This works especially well when the periodicity of the cycle is
long enough and the oldest data is cold enough that the chance it cannot be
archived clean or graduated to long term lower priority access before it
needs to be actually discarded from the system being pruned is vanishingly
small.
On the other hand, if the information grouping of the company for a
particular system is month, then it makes sense to drive your information
life cycle off months. I'm actually a bit curious what is driving the need
for it being "exactly" two years.
As a general principal I encourage the notion that folks think strategically
about how they will handle the disposition of data in the long haul as a
priority concern when the system is designed. There is a strong
anti-correlation between this type of thinking and the type of system that
runs just fine for a few months or a few years after it starts accumulating
data and then hits a crisis of size that may be difficult to escape like a
box canyon during a gully washing cloudburst. It should also drive policy
concerns about how to handle manipulation of old data when the incumbent
schema is improved, but I've already written too much and I'll just leave
that giant topic dangling.
Regards,
mwf
_____
From: oracle-l-bounce@xxxxxxxxxxxxx [mailto:oracle-l-bounce@xxxxxxxxxxxxx]
On Behalf Of rjamya
Sent: Thursday, April 19, 2007 10:17 PM
To: smishra_97@xxxxxxxxx
Cc: oracle-l@xxxxxxxxxxxxx
Subject: Re: Delete Data older than exactly 2 years
<snip>
Month to month cleanup is always better than on a random date in the last
week. What do you think?
<snip>
Other related posts:
