Yes, sounds correct. You can also get "in-process" scaleout on a multi-core machine by asking UIMA-AS to create multiple instances of your AE (assuming it is thread-safe) , e.g. <scaleout numberOfInstances="8"/> If you use jconsole to monitor the broker to verify the number of consumers, note that UIMA-AS creates 1 extra listener per deployment, for a total of 9 per deployment in this example.
~Burn On Wed, Jul 30, 2014 at 11:09 PM, Swirl <[email protected]> wrote: > Swirl <lriwswirl@...> writes: > > > > > Hi, > > I am trying to use Uima AS to deploy one of my UIMA Aggregate AE. > > I wish to deploy it as a Uima AS Primitive, i.e. I do not need the > delegates > > of this AE to be scaled out. Instead I just want to scale out the Uima > > aggregate AE only. I want to deploy the AE in multiple nodes. > > > > > OK, some answers from myself... > > I simply copy all required jars (of my AE) and descriptors to all machines > that I want the AE deployed. > I start the broker on one of the machines (it should be possible to use a > dedicated machine to act as the broker). > I do a "deployAsyncService" on all the machines, specifying the "brokerURL" > as tcp://<brokerMachine>:61616. > > > To test, I do "runRemoteAsyncAE" on a separate machine containing the input > files to be passed to the AE, again specifying the "brokerURL" as > tcp://<brokerMachine>:61616. > > From my logs on all the machines, I can see that some CASes are distributed > to all the machines with the deployed services. > > Are the above steps correct? > >
