User Tools

Site Tools


Experiment execution

Code SREM-2
Responsible Leandro Navarro
Components testbed server, testbed node


Once the slice is started, the experimenter should be able to start/control/stop the experiment. This means executing commands to start applications, trigger events (failures), etc.


Assuming a slice is configured (resources are allocated), slivers are active and thus ready for launching an experiment.

This requirement is about the necessary steps to control/coordinate the life cycle of an experiment (start, stop, collect results, cleanup) involving any number of slivers.

We can borrow most of the terminology from PlanetLab if not suggested otherwise.



Researchers may want to manage its experiment in different ways depending on the type of experiment, the type of software used, their preferences regarding preferences on automation or detailed control, etc.

Tools for running an experiment range from simple tools like vxargs to more complex tools with features for the automation of experiments such as Plush,Nebula or OMF.

Open discussions

What services/hooks should be included in the testbed management system to support these tools (e.g. a daemon may be required to run at each slice)


Confine is quite similar to Planetlab but over a set of nodes inside community networks.

  • I'd suggest to start with no specific support (unbundled experiment management) other than ssh access to slivers (leaving the users to choose their own tools (e.g. vxargs)
  • I'd suggest to investigate how tools like OMF or Plush fit into the characteristics of community networks and provide the necessary support in the CONFINE testbed (e.g. APIs on the server, perhaps SFA-like services) or explore the necessary changes required in our specific environment (less reliance on a stable infrastructure).
requirements/experiment-execution.txt · Last modified: 2012/01/31 09:46 by leandro