[omniORB] g3 and gif files

Alberto Squassabia alsq@alsqhpw.cnd.hp.com
Mon, 10 May 1999 08:41:10 -0600


Haven't tried it on omniOrb specifically, but octets are unmarshalled
and should fit your purpose with the best efficiency.  In my experience
the bandwidth of a bulk transfer varies with the size of the
transfer.  Small chunks (<1K) are dominated by protocol overhead,
and throughput is disappointing.  Large chunks (1M) do much better,
I reached close to 8 Mbit/sec (YUP!) payload b/w.  Keep in mind, though,
that my number was measured on a private ethernet (eg., no traffic
but mine).  Mileage will vary with (1) size of each transfer, (2)
background network usage, and (3) the phase of the moon (eh, eh).

[Re: [omniORB] g3 and gif files] grohmann@ike.uni-stuttgart.de wrote:
>
>
>
> andrew.brown@bt.com wrote:
>
> > Hi,
> >         Does anyone know how i would go about passing a .gif file and a .g3
> > file through a CORBA interface (specifically OMNIorb). Any information or
> > ideas on this would be much appreciated.
> >
> > Thanks.
> >
> > Andrew Brown.
>
>  Andrew,
>
> I am not sure whether it is a good idea to pass files (especially large
files)
> via CORBA. OmniOrb uses a default maximum message size of 2 MB, and my
> impression is that any CORBA method invokation which passes more than, say
100
> kB, is a misuse of CORBA.
>
> In our implementation, we separate between the control flow, i.e. information
> about the file being exchanged, and the data flow. The control flow is
handled
> by CORBA, and the data flow is facilitated by either a distributed file
system
> or FTP.
>
> Cheers, Axel
>
> --
> Axel Grohmann
>
> Department of Knowledge Engineering and Numerics, IKE
> University of Stuttgart
> Pfaffenwaldring 31
> D-70550 Stuttgart, Germany
> Phone: 0049 711 685-2130
> Fax:   0049 711 685-2010
>
>
>
>>>-- End of excerpt from grohmann@ike.uni-stuttgart.de -----------------<<<



--
Alberto Squassabia  (970)898-7705  alsq@cnd.hp.com  HP TMD
"   We are Pentium of Borg. Division is futile. You      "
"   will be approximated. (ca. AD 1995)                  "