Copy of 24

1
a presentation by W H Inmon
2
DSS
OLTP
There are two worlds that are very different
Both need to be managed, but there are
entirely different objectives and approaches
3
the nature of data warehousing is growth -
- historical data - detailed data - summary
data - unknown requirements
4
managing the warehouse
small warehouses - - easy to manage -
cheap - no user complaints
large warehouses - - a challenge to manage
- expensive - user complaints - performance
is difficult to achieve
5
a lot of people use the data warehouse
6
the information being managed by the data
warehouse becomes mission critical
7
data warehouse has become the foundation for
applications of the 1990s
8
managing the data warehouse is quite different
than managing other systems resources
9
response time
OLTP
in the world of transaction processing, response
time was the most important facet of the system
to manage
10
DSS
in the world of data warehousing, the volume of
data is the most important facet of the system to
be managed
11
DSS
the volume of data has a profound effect on -
- expenses - performance - availability
- accessibility
12
as data warehouses grow, the data divides into
two classes - - frequently used
data - infrequently used data
dormant data
dwmgmt01
13
actively used data is kept in high
performance disk storage inactively used data is
moved to near line or secondary storage
dwmgmt01
14
the problem with dormant data - - it is
GROSSLY wasteful - it is expensive - -
- it hurts performance for everyone by
getting in the way of actively used data
dwmgmt02
15
so how do I find dormant data?
dwmgmt02
16
SQL
SQL
SQL
SQL
the data that queries are accessing is determined
dwmgmt03
17
SQL
SQL
SQL
SQL
many other useful pieces of information are
gathered as well -
- top ten accessed tables - average query
response time - longest running queries - how
many duplicate queries - times of peak usage -
are data marts appropriate?
dwmgmt03
18
SQL
SQL
SQL
SQL
what can you learn from an monitoring your
activity?
19
near line storage
where do you put dormant data?
20
what data marts should I create and how should I
structure them?
21
does my processing workload warrant an
exploration warehouse?
22
iterative design issues -
where should I summarize?
where should I create a fact table?
where should I aggregate?
23
SQL
SQL
SQL
SQL
but there are other aspects of the
environment that are also important...
dwmgmt03
24
such as - What is going on inside the data
warehouse environment RIGHT NOW!
25
immediate answers -
- what queries are running now?
dwmgmt03
26
immediate answers -
- what SQL is in execution now?
dwmgmt03
27
immediate answers -
- how does todays execution compare to
previous executions?
dwmgmt03
28
immediate answers -
whose queries are these anyway?
29
immediate answers -
is it safe to cancel a query that has been
in execution a long time?
30
immediate answers -
the ability to measure in flight data
31
immediate answers -
are my service level agreements being met?