We are using Mule in real-time situation to connect several systems
together. As you stated,
half-second is a very long time because we typically measure in
milliseconds.
We have also found WS* to be slow for real-time transactional systems.
I'm sure you know
this already, but end-to-end performance is not just dependent of a
protocol. The structure
of the data and how they are created as well as how
business/presentation layers handle the
data impacts performance. There are several products and experiments to
boost WS*, and
it nice to keep track of them.
As of now, WS* may be used in situations with small data and few
requests, and we do
actually use them too.
FYI:We actually are vendor neutral and is not commited to just using
Mule I am just using it
because it's open source and let us easily experiment and also
because it allows me to more
easily create custom solutions.
Cheers,
H.Ozawa
Bill Barr wrote:
> Regrettably, that is something I cannot do for several legal reasons. I
> should have qualified my statements that we performed our tests under a
> typical, real-time load. However, I will see if I can post the actual
> results. As for performance requirements, that is determined by the
> impatience of a customer using our site. :) Our usability studies
> indicate that every half-second of latency at the user interface is
> very, very expensive.
>
> --
> email: [EMAIL PROTECTED]
>
>
