Thread: Are there any performance penalty for opposite edian platform combinations....

Are there any performance penalty for opposite edian platform combinations....

From
"Guoping Zhang"
Date:
Hello,

I apologize that if the similar questions were already asked and answered
before.....
Here is a go:

a) If we have application clients running on a Solaris 10/SPARC box and
database server running on a Solaris10 X_86 box; further, we have a few
tables, in which we split an integer type of field into several our own
defined bit map segement, upon them, we have a set of logic operation
implemented in our applications, MY question is, could the different edian
scheme (SPARC is a big edian and X_86 is the opposite) possibly slow down
the applcation?

In fact, it is a general question that "Is it a good practice we shall avoid
to run application server and database server on the platform with opposite
edian? or it simply doesn't matter"?

b) Same direction for the question, if using slony-1, if master server is
running on a big edian host but slave is running on a small edian host, are
there any performance loss due to the edian difference?

Thanks in advance for your opinions.

Regards,
Guoping Zhang


"Guoping Zhang" <guoping.zhang@nec.com.au> writes:
> In fact, it is a general question that "Is it a good practice we shall avoid
> to run application server and database server on the platform with opposite
> edian? or it simply doesn't matter"?

Our network protocol uses big-endian consistently, so there will be some
tiny hit for little-endian machines, independently of what's on the
other end of the wire.  I can't imagine you could measure the difference
though.

            regards, tom lane