>
> Is there a significant benefit of using connection pooling over a
> single threaded model wrapped in an actor?
Do you mean maintaining a connection per thread, or maintaining one
connection in an actor, or?
A connection per thread would only be efficient if the actors are long
lived -- SimpleDB's HTTP setup time is 1-2 seconds without an
existing perstistent HTTP/1.1.
However, that is still a -long- time to wait when setting up a new
actor (eg, tacking on 1-2 seconds to the first page load, etc).
Wrapping one connection in one actor would disallow parallelism, which
is where SimpleDB shines. Maintaining a pool of actors would be
replicating what commons-pool provides.
Hopefully I didn't misinterpret your query.
-landonf