[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [dist-obj] CORBA implementation ideas
Mark R. Thomsen wrote:
> >Why does Doug Schmidt's considerable investigation into this area seem to
> >contradict your claims?
> Pointer, please. I am unfamiliar with this reference. I will look and see
> what I can deduce.
for Doug Schmidt's and Andy Gokhale's paper on analyzing the
issues with IIOP performance and applying a systematic
set of optimizations
to the Sunsoft reference implementation of IIOP which dramatically
improved overall performance, esp. for dynamic invocations.
They've got other papers on this theme since which I haven't
read, but which also look relevant, eg.
Look, I think the basic theme is that marshalling complex
data types has a cost, no matter what you do. Obviously
you can go faster dumping raw byte arrays over a socket connection,
and sometimes that's the right thing to do. But if you believe
in OO as a vital systems-organization tool, I think it's reasonable
to try to used typed message exchanges (ie marshalling) insofar
as you can tolerate the hit. Doug's work indicates that if you
really know what you're doing, you can reduce the hit, and
get far, far better performace from a marshalling protocol
(eg IIOP) than most of us dodos can typically do, which is
why we bail out for raw sockets.
(BTW, it's a serious mistake to equate Doug Schmidt with CORBA.
A visit to his site takes you through a whirlwind tool of
high performance webserver research, Java, ACE, med. applications,
patterns, networking, ... As I've said before on this list,
I always get depressed when I visit there. Where the *bleep*
does he get the time to be so darn productive??? I'd be
ecstatic with 5% of his output!)
Oh, once you hop over there, check out:
Apparently CORBA research really can pay off, literally. Doesn't
look like the good Dr. Schmidt is hurtin' any :-).
(I'd post a jpg of my car, but who wants to see shots
of a beatup old '91 station wagon loaded with
ice-cream stained kiddie-seats?)
To unsubscribe and other requests, mailto:firstname.lastname@example.org
Archives, FAQ, etc. http://www.distributedcoalition.org/mailing_lists/