The ATLAS TriggerTool: one number defines it all

2 November 2009

Shifter at work at the trigger desk in the ATLAS Control Room



"What’s the current supermaster key?" is a question you’re very likely to hear should you find yourself in the vicinity of the trigger or run-control desks of the ATLAS control room. The answer to this question, a single integer, is enough to tell you exactly how the ATLAS trigger is configured at that point in time, right down to the last configurable parameter of the trigger software. Put another way, it tells you the trigger menu that is being used to take data.

As many people know, the ATLAS trigger system must reject 99.9995% of all events, while maximising the efficiency for those few events that show signs of interesting physics processes. This is necessary in order to reduce the 40 MHz bunch crossing rate to a manageable 200 Hz data storage rate. Which events are kept depends on the presence of their detector "signatures" in the trigger menu that is being used. The menu must cater to all possible tastes when it comes to physics processes making it a comprehensive, and complicated, list: by way of a first course, up to 256 distinct trigger items (like finding two 20 GeV jets, or a single 25 GeV muon) are offered at Level 1. This is followed (and refined) by hundreds of possible signatures at the next level, namely the High Level Trigger (HLT) software, corresponding for example to some complex tau or b-physics processes.

Once prepared by the menu group, menu files for different operating scenarios are uploaded to an Oracle database at Point 1 (the Trigger Database or TriggerDB for short). There they are assigned their ubiquitous supermaster key. For those familiar with database design, this key is the "primary key" of the top-level table in a hierarchy of tables associated via many-to-many links. As such, it uniquely picks out the entries in the tables below that, once put together, completely define the menu.

The menu files are uploaded to the TriggerDB thanks to the TriggerTool, the graphical user interface to the database (http://cdsweb.cern.ch/record/1197160). This application, currently developed by the Manchester and DESY groups, comprises many thousands of lines of java code. It has been carefully designed to adhere to two principles:no records are ever duplicated in the database and the history of all menus can be maintained. This means that a certain tau signature  that is used in many menus has exactly the same definition, down to the last selection cut, in every menu in which it appears. This avoids having any ambiguity about the trigger configuration. It also means that the menu defined by supermaster key 614 will be exactly the same in ten years’ time as it was yesterday when it was used to take data, thus providing a future reference to the trigger conditions of any run.

For the person who is on shift at the trigger desk, the TriggerTool is the familiar face of the wider configuration system that encompasses not only the TriggerDB but also the behind-the-scene software that quietly deploys the configuration data for a selected supermaster key to the thousands of HLT nodes and the Level 1 hardware, thereby realising the desired menu. For a shifter, the TriggerTool is one of the most important applications at his or her disposal, allowing them to visualise any menu in graphical form or perhaps compare the menu of the current supermaster key with the previous one. In addition to providing the graphical display, the Tool allows changes to be made to the configuration (and hence new supermaster keys to be generated) that can be quickly used online. For people in search of trigger configuration, information previously used in data taking, a run-wise search for supermaster keys is available at http://trigconf.cern.ch, which uses the TriggerTool as a back-end to create a web page display of the menu.


Screen shot of the TriggerTool graphical display (click for full image)



The TriggerTool also has applications in the offline world. It is used for example to "replicate" certain supermaster keys that define trigger menus used in Monte Carlo production from the central database to special database files (in SQLite format). The supermaster keys in those files can then be installed for use at the Tier 2 production sites. In addition, it is also possible for people to copy menus of interest from the Point 1 database to their own private databases, from where they may rerun the trigger and study the effect of making changes to the menu they may be interested in.

The TriggerTool and the database configuration system have been in use at Point 1 for over two years, ever since the M4 commissioning run in 2007.  Since then, the term supermaster key has become embedded in ATLAS trigger vocabulary. It will probably be unavoidable for anyone performing a physics analysis in the future, and we hope that our tools, which continue to improve, will allow them to make sense of the trigger conditions under which their data was taken.

 

Paul Bell

University of Manchester