Speaker
Moritz Kretz
(Ruprecht-Karls-Universitaet Heidelberg (DE))
Description
With the installation of the Insertable B-Layer in 2014 the Pixel Detector of the ATLAS experiment has been extended by about 12 million pixels. Scanning and tuning procedures have been implemented by employing newly designed read-out hardware which is now able to support the full detector bandwidth even for calibration. The hardware is supported by an embedded software stack running on the read-out boards. Compute-intensive operations that are necessary during the calibration process are performed on a small cluster of commodity servers, which are included in the ATLAS TDAQ framework. In order to exploit the potential of this heterogenious data processing architecture to significantly speed-up detector calibration, timing and performance of the various software components is critical.
We present a timing model including the hardware and software components and procedures that are involved in the calibration process. By instrumenting these components at critical locations we are then able to accurately evaluate the timing of a calibration scan, analyze the performance, and eventually locate and eliminate bottlenecks in the chain.
Author
Moritz Kretz
(Ruprecht-Karls-Universitaet Heidelberg (DE))
