Error Table Population...

[him121]

hi..
all
here we are developing Fact job in that.
we have to populate Error Table for each and every lookup..

now the situation is we have 4 different Error Table..
E1,E2,E3,E4...

and we have 24 lookups..so..for all the 22 lookups we have to populate the Error Table..

now....in my job ..i m doing 6 lookup in each Transformer Stage..

and i am checking ErrorFlag...for all the 6 lookup in that....stage..
and ..taking that.column..up to last transformer....

design looks like..

Transformer1 + 6 lookups + 6 ErrorFlag columns..
Transformer2 + 6 lookups + (6+6=12) ErrorFlag columns..
Transformer3 + 6 lookups + (6+6+6=18) ErrorFlag columns..
Transformer4 + 6 lookups + (6+6+6+6=24) ErrorFlag columns..

in Transformer 5 we have now all the 24 columns..
then i am populating E1,E2,E3,E4...error tables...

can any one tell me..taking 6 columns....
upto last transformer..will affect the performance..

[sumitgulati]

Him, I don't think there will be any noticable performance drop because the data contained in those columns is not huge. It is just a Flag value ("Y" or "N").

-Sumit

[ray.wurlod]

I agree with Sumit's assessment. Thirty Char(1) columns requires hardly any processing at all. A column that is passed through a Transformer stage with no derivation requires zero machine cycles (and therefore has zero cost); inspect the code generated by compiling a Transformer stage to convince yourself of this.