> > No offense but... has anyone used RMI for real systems?
> I've tunneled CORBA
> > traffic through HTTP and run IIOP over SSL - in working,
> real-life systems
> > ;-)
> > Well, I'm a RMI newbie (I've only written one application
> for performance
> > testing) so I'd really like to hear about your
> real-world-RMI experiences!
> Almost. I worked on a production CORBA system with
> applet client
> and custom server for intranets (1-5000 users, no tunneling
> or anything).
> We did a trial conversion to RMI & EJBs, and found the
> performance of RMI
> to be significantly better than CORBA, though that may be due
> to the fact
> that we made heavy use of CORBA Anys. Anyway, we didn't end
> up converting
> at the time (a few months ago) because a major client
> demanded Macintosh
> support, and while we could use a Java 1 client and Java 2 server with
> CORBA, we could not find a good EJB platform that would allow
> that (and
> Mac doesn't have a Java 2 VM yet).
CORBA Anys are sloooow... don't (mis)use 'em if don't have to;-)
1) if you have a fixed number types that you expect to pass you should
really use unions instead... much faster.. and better type safety...
2) if you're only sending object references, well a general CORBA::Object
will do the job...
3) you might also want to look at value-types (if your ORB supports 'em)
If you compare performance for SUN:s implementation of IIOP (in JavaIDL) and
their implementation of RMI/JRMP you'll see that RMI/JRMP is faster.
However, SUN doesn't have a very good track-record for building CORBA
ORB:s... I you want speed, get hold of a decent CORBA C++ ORB (fancy
implementations can use stuff like shared memory transport for intra-process
communication on one host) or use pure socket communication...
Personally I'm do not think that "performance" should be the only (or even a
very important) criteria for choosing a particular technology or
implementation (there are loads of other aspects which are more interesting
e.g. stability, scalability, ...).
As for running RMI-over-HTTP (or IIOP-over-HTTP) I think it's not the right
way to go anyway... I was just curious to hear if anyone had actually used
it. Stuff like SOAP or OMG:s SCOAP effort might be better approaches, even
though SOAP is 30-60 times (a.k.a. "significantly") slower than both
RMI/JRMP and JavaIDL/IIOP...
Regards,
Per
--
--------------------------------------------------------------
To subscribe: [EMAIL PROTECTED]
To unsubscribe: [EMAIL PROTECTED]
Problems?: [EMAIL PROTECTED]
