Hello,
Indeed, since the server discovery protocol, based on UDP multicast, will not work over the internet, you will need to disable it and provide the server address and port in the node's configuration. For this simply add the following to its configuration file, located at JPPF_NODE_INSTALL/config/jppf-node.properties:
jppf.discovery.enabled = false
# must be an (public) address reachable by the node
jppf.server.host = host_or_ip_address
jppf.server.port = 11111
Our documentation provides more details on the
node configuration.
To answer your second question, yes, nodes can be added (or removed) dynamically to the grid, and taken into account immediately for the jobs being processed. How the jobs and their tasks are distributed to the nodes can be directly influenced by your
load-balancing settings, as well as by the
jobs service level agreement (SLA).
Additionallly, I would like to point out that the latest unstable version
JPPF 4.0 alpha 2, released yesterday, brings a set of features specifically intended to support a BOINC-like volunteer computing model. In particular, I invite you to read about the new
offline nodes feature which, when combined with the
"idle host" mode (i.e. cpu scavenging), should provide you with a powerful alternative to what already exists.
I hope this helps.
Sincerely,
-Laurent