Page History

There are three different ways to get quasi real time information about data taking. They offer different pros and cons and the final choice about which method works best for a given experiments is left to the users.

The approaches are:

1. AMI (Analysis Monitoring Interface) processes running on the instrument dedicated monitoring nodes.
2. Shared memory applications running on the instrument dedicated monitoring nodes.
3. Custom applications running on the experimental hall fast feedback queues and reading the data from disk.

AMI

AMI can display the raw/calibrated detector data in real time and it can calculate and display some simple quantities without writing any code.  Analyses are configured through a GUI with just a few mouse clicks. A guide to using AMI can be found here.

...

Cons: non trivial augmenting capabilities through additional C++ code.

...

This method of analysis reads the detector data from a fast storage layer called fast feedback (FFB), dedicated to the running experiment, and runs a prepared analysis. Any framework which can read XTC files can be adopted to run as a fast feedback application. The most common framework used with this method is psana.

Pros: trivial to setup; independent of the DAQ; ability to run the same code online and offline.

Cons: not quite real time (<5 minutes); may compete with the HDF5 translator if the translation services are run on the FFB storage layer.

Shared Memory (Experts only!)

...

Cons: if not using pre-existing software, requires very large time investment from both LCLS and the user.  competes with AMI resources (monitoring nodes); requires setting up the shared memory servers on the DAQ side.

Fast Feedback

This method of analysis reads the detector data from a fast storage layer called fast feedback (FFB), dedicated to the running experiment, and runs a prepared analysis. Any framework which can read XTC files can be adopted to run as a fast feedback application. The most common framework used with this method is psana.

Pros: trivial to setup; independent of the DAQ; ability to run the same code online and offline.

Cons: not quite real time (<5 minutes); may compete with the HDF5 translator if the translation services are run on the FFB storage layer.