SOAP slower than Java RMI?

[Paul Lessard]

I saw on another post that SOAP can be 10 times slower than java RMI. Is this true in general, or is that a cieling?

[Glenn Murray]

Is SOAP slower than RMI? Well, yes, of course. Ten times slower? That depends. SOAP is designed for using XML (ASCII) and language independence, not performance.

[Aleksey Matiychenko]

I am not sure how you make that comparison.
Both SOAP and RMI make network calls so there is a performance issue there.
SOAP may be slower since you may need to parse XML and create object but it also allows for asynchronous messages so you don't have to wait for results to complete. I know that RMI has callbacks but those are extremely slow.
So as always in this field the answer depends on what your are trying to achieve.

[Glenn Murray]

Not "may be slower". It will be slower, for the reasons Aleksey pointed out. The original question is about RMI, not asynchronous messaging or "network calls".
In an abstract, apples-to-apples comparison, if you use SOAP to do remote procedure calls with Java on both ends, then RMI will be faster every time.
This is not to say that for any particular application RMI is superior; as Aleksey also pointed out, it all depends.

[Andrew Sweeney]

I think that the main point of SOAP is being lost, in that it allows inter-system communication and RPC that is language independent. Sure, RMI will always beat out SOAP. Then again, RMI will usually beat CORBA with an all-Java architecture. (I believe, anyway) Performance is A goal, but it isn't the end-all where SOAP is concerned.
Cheers,
Andrew

[Andy Rodriguez]

More on Aleksey's point , the IBM article on Performance testingdetails the "points of failure" ( refer figure 2).
And in terms of numbers , the article on S-O-A ( next hip thing according to gartner )vincihas references to how SOAP , RMI fared on a head to head test
[ January 22, 2002: Message edited by: Andy Rodriguez ]