“Our society has become
so used to super-sizing we take big for granted. Big cars, bigger burgers and
now unnecessary biggest BI systems. We have to redefine normal. It’s not normal
to not delete your PSA so you become the largest BI environment, it’s not normal
to supersize everything. That it’s unhealthy to be big, but its unhealthy and
extremely unhealthy to be fat, podgy and believing that big is the secret to
success.
Lose weight, don’t waste
energy, build a regimen of global methodology, deploy FEDW Architecture, use more automated modeling, lower data
throughput pressure, conduct annual health-checks, watch your dirty data. A
six-pack does not mean building six podgy BI environments across the planet.
These health rules all still stand, but BI Value doctors have added new
recommendations and caveats to the list. It’s not normal not to optimize your
InfoCube models and just filling your BW with more and more data. Its not
normal to put all your obese cubes into a BW Accelerator, as it will not be normal
to put all your obese cubes now to HANA. Just like it’s not normal to always
drive and never walk, so too it’s not normal to simply keep putting more and
more data without running a BI Health Check to ensure that your data, ODS,
Cubes and BI are lean and mean and built on muscles and not just pompous fat.
Executive, budgets and corporations get numb to the concept of diet and
exercise” Dr. BI Valuenomics.
Build a global regimen, lose weight, don’t waste energy(smoke), use less
unnecessary data (Salt), lower data throughput pressure (blood pressure),
conduct frequent automatic remodeling (exercise), watch your dead data
(cholesterol), don’t install the wrong application because the salesperson told
you so (marry the wrong person). These health rules all still stand,, but Dr
Scientific Principles has added new recommendations and caveats to the list.
So let’s check if you understand the standard health rules.
a.
Build a global regimen: Corporations can no
longer survive in their local ponds. The flat world converts each small company
into a global multinational. Similarly BI projects are no longer built locally.
It is a mixture of on-site, off-site and offshore. In order to ensure that all
the teams work harmoniously the first step in all BI initiatives is to build a
global methodology of standards and processes at a minimum. This will not only
save your tactical life, but contribute to your strategic living.
b.
Lose weight to enhance performance: It has been adequately
established that the average DSO or Cube in North America may contain 40 to 60%
redundant objects. This is great if you collect data but pretty stressful for
performance and support SLA’s (Service Level Agreements). So get an automatic
audit of your Cubes and/or remodel your DSO’s to optimized Cubes so you perform
at your capabilities. As a BW moves to the BW Accelerator phase the Cost of
Ownership for each data element goes up considerably, due to the high cost of
HW and support. By decreasing the BW size by 40 to 60% we have been able either
decrease both the initial cost and annual support cost by the same proportion
or allow 40-60% more cubes into the same BWA.
c.
Use Less Unnecessary Data: Our research indicates
that an average north American ODS or Cube can contain anywhere from 20 to 60%
data elements that are redundant. By
identifying and eliminating these elements we consume lesser system respurces
and meet the objectives of a.
d.
Lower data throughput pressure: Simply by building a
cube on top of an ODS we have been able to reduce query times from 500-700
seconds to under 20 seconds. By eliminating all the redundant objects we were
further able to reduce this to under 13 seconds. Useless data does much more
than occupy space. It clogs the system resources and slows data flow along with
Query performance.
e.
Conduct frequent remodeling: Just as a car built in
1911 could not compete with the cars of today, nor a car built in 1997 have all
the capabilities of ones built today. Similarly the cubes modeled in 2001
cannot be used as the information demand and priorities change. Do not try to
do this manually as that can not only be costly but also extremely ineffecient.
Go for the automatic route and finish this task in a few hours per cube.
f.
Watch your dead data: Our data warehouses are
full of test elements, KPI’s and queries. This is dead-wood sittin gin your
racing lance. Flush your system frequently so the engines perform optimally
g.
Don’t install the first application that
comes your way: Just like don’t fall for the first person, or buy the
first car you see. So too don’t decide on your final application just because
someone told you so. We are finding more and more that the sheer
competitiveness in the BI environment creates salespersons who truly don’t
understaand their own products. Get the opinion of a trusted BI Business Value
Architect prior to finalizing decisions.
So think scientifically and use your head to save your heart.
No comments:
Post a Comment