Warning!
PrerequisiteFor this tutorial we assume that you know how to launch PLASMA Lab and run a local experiment. If you don’t know how to do that please refer to the other tutorials.
You also need to run PLASMA Service. Please refer to the download page to learn the different methods available.
Setup a PLASMA Lab distributed experiment
When setting our experiment we didn’t used the execution panel which was set to local by default.This panel allows to configure 3 different executions mode:
- A local single thread mode that is used by default when the selector is on local and the number of threads is equal to 1.
- A local multithreaded mode when the selector is on local and the number of threads is at least 2.
- A distributed mode that waits for worker to connect to the experiment.
Before clicking on Start, we now select the distributed mode. Then we can set the port on which PLASMA Lab is listening. We can also configure the size of the jobs that will be ask to each worker by changing the Batch parameter.
Execution panel
When we click on the Start button PLASMA Lab will wait for workers from PLASMA service instances to connect and it will then send tasks to these workers.
Setup PLASMA Service
PLASMA Service interface
PLASMA Service lets you define a host address and a port on which PLASMA Lab is listening. You can also set the number of threads that you want to run on this instance of PLASMA Service.
Once PLASMA Service is setup we click on the connect button. PLASMA Service status now switch to waiting or immediately start computing if PLASMA Lab has already assigned tasks. Once our experiment is completed, PLASMA Service disconnect itself.
If you want to know how PLASMA Lab and PLASMA service interact together, go to our architecture page.
Alternatively you can run PLASMA Service in a console. This might be useful if you intend to run it on a distant machine.
Important!
We used Igrida to test the scaling of our distributed algorithm. Check our experiment here.In the next step of our tutorial you can learn how to use the optimization panel.
