Database Reference
In-Depth Information
ETL
ELT
ETLT
pendent of and
outside the
database and
is reusable.
• Works very
well for granu-
lar, simple
function-orien-
ted transforma-
tions that do
not require any
database calls.
• Can run on
SMP
(
Sym-
metric Pro-
cessing
) or
MPP hardware.
• Keeps all data
in the RDBMS
all the time.
• Is parallelized
according to the
data set and
disk I/O is usu-
ally optimized at
the engine level
for faster
throughput.
• Scales as long
as the hardware
and RDBMS en-
gine can contin-
ue to scale.
• Can achieve 3x
to 4x the
throughput rates
on the appropri-
ately tuned
MPP RDBMS
platform.
• Requires high-
er processing
power on the
ETL side.
• Higher costs.
• Complex trans-
formations that
would need
reference data
would slow
• Transformation
logic is tied to
database.
• Transformations
that involve
smaller volume
and simple in
nature would
not gain much
benefit.
• Would still
have a part of
the transform-
ation logic
within data-
base.
Risks
Search WWH ::
Custom Search