Very interesting indeed, I had the same initial reaction that it was
simplistic.
I think that relying on the DNS roundrobin trick is actually pretty smart
and "low-tech". I find their recommendation of NFS'ing their critical
persistent structures really "low-tech" and smart (I was heading the same
way btw).
The naming I had never really sat down and thought about it but I think it
boils down to a trick in the bind that multicasts to all registered copies
in the cluster. JNDI event Bind is thus a new thing we need to either
multicast either just Event Propagate. I think that is fairly straight
forward.
I understand the "stateless" clustering, that is fairly trivial and taken
care of by the above trick combo. The Entity stuff they don't really
explain the cache, but I suspect they don't really deal with it? (don't
know). For the stateful stuff I am not sure I understand. The transaction
means that the stuff becomes persistent in their mind (what do they mean by
"relook the home and reapply the transaction"). I had something similar in
jboss1.0 where the stub was already clusterable and would survive the crash
of a stateful server given that the instance was passivated (on disk). It
is actually fairly trivial (couple of lines in Proxy stuff go check it out).
Other than that, I had this strange feeling that sometimes "low-tech" is
good. See I have a phD in applied stuff, and I know everything breaks all
the time and nothing works... so something simple is good. It reminds me of
the time the US air force captured this Russian aircraft and found old
"bulb" electronic stuff inside. At first they were laughing their heads
off, thinking they were starting only and then they realized that these did
not break under EM radiation :). Low tech is good sometimes.
So let's not assume that WL did this without thinking. Maybe there is
little more to it and then we will look at the real problems in the caches
;-)
regards
marc
> -----Original Message-----
> From: [EMAIL PROTECTED]
> [mailto:[EMAIL PROTECTED]]On Behalf Of Rickard �berg
> Sent: Monday, July 17, 2000 11:00 PM
> To: jBoss Developer
> Subject: Re: [jBoss-Dev] Clustering of JBoss
>
>
> Hi!
>
> Adi Lev wrote:
> > Look on WebLogic clustering logic from BEA,
> > read the documentation, here is a url for the documentation:
> > http://www.weblogic.com/docs51/intro/intro_admin.html
>
> Interesting reading. My initial reaction is that it does seem a little
> simplistic, and don't think it will be very difficult to do something
> more intelligent. We will see.
>
> /Rickard
>
> --
> Rickard �berg
>
> @home: +46 13 177937
> Email: [EMAIL PROTECTED]
> http://www.telkel.com
> http://www.jboss.org
> http://www.dreambean.com
>
>
