"Network of Brokers" and persistence

[Steve Cohen]

I am in the phase of imagining what using ActiveMQ to design a wrapper 
around a legacy process would look like, and reading the book, which I 
have bought.  I should say that I am impressed so far with ActiveMQ and 
the mapping of what it does with what I am trying to do seems very good.

I am trying to understand the relation of persistence to the "network of 
brokers" concept.  In a single standalone broker deployment, it's 
simple.  You either enable persistence of one flavor or another, or you 
don't.

But what does this look like in the "network of brokers" concept?  There 
is something appealing in this model to my situation, of deploying a 
server-side application in which each instance has an instance of the 
broker embedded within it, but what are the consequences in terms of 
persistence?  Would there just be one persistent store, with a suitable 
backup arrangement?

Please help me untangle the consequences of these two concepts, which 
are starting to boggle my mind a bit.