By alexismp on Oct 03, 2011
The GlassFish Cloud demonstration in the JavaOne Technical Keynote touched upon many different parts of PaaS, including elasticity - the key to better resource utilization, one of the key motivation for cloud.
To understand further how auto-scaling is done in GlassFish, this page lists similar instructions to the overall PaaS demo (using a recent build of GlassFish 4.0), only focusing on the specifics of elasticity with the notions of alerts and alert actions.
In this case the demo uses
asadmin create-alert and
asadmin add-alert-action to define a threshold value and associated action based on memory usage (just one possible metric among others). You can then artificially increase or decrease the memory load and thus scale up (dynamically provision and add instances to the cluster) or scale down the entire service (shut down and delete instances). You can also observe all of this using the monitoring charts included in the PaaS console.
For more details on elasticity, check out material for "JavaOne 2011 - Demo for AutoScaling in the Cloud Technical Session (26462)" as it becomes available.