|
The High Level Trigger of the ATLAS experiment relies on the precise knowledge of the position, size and orientation of the luminous region produced by the LHC. Moreover, these parameters change significantly even during a single data taking run. We present the challenges, solutions and results for the online luminous region (beam spot) determination, and its monitoring and feedback system in ATLAS. The massively parallel calculation is performed on the trigger farm, where individual processors execute a dedicated algorithm that reconstructs event vertices from the proton-proton collision tracks seen in the silicon trackers. Monitoring histograms from all the cores are sampled and aggregated across the farm every 60 seconds. We describe the process by which a standalone application fetches and fits these distributions, extracting the parameters in real time. When the difference between the nominal and measured beam spot values satisfies threshold conditions, the parameters are published to close the feedback loop. To achieve sharp time boundaries across the event stream that is triggered at rates of several kHz, a special datagram is injected into the event path via the Central Trigger Processor that signals the pending update to the trigger nodes. Finally, we describe the efficient near-simultaneous database access through a proxy fan-out tree, which allows thousands of nodes to fetch the same set of values in a fraction of a second.
|
|