The upgrade of the ATLAS High Level Trigger and Data Acquisition systems and their integration

The Data Acquisition (DAQ) and High Level Trigger (HLT) systems that served the ATLAS experiment during LHC's first run are being upgraded in the first long LHC shutdown period, from 2013 to 2015. This contribution describes the elements that are vital for the new interaction between the two systems. The central architectural enhancement is the fusion of the once separate Level 2, Event Building (EB), and Event Filter steps. Through the factorization of previously disperse functionality and better exploitation of caching mechanisms, the inherent simplification carries with it an increase in performance. Flexibility to different running conditions is improved by an automatic balance of formerly separate tasks. Incremental EB is the principle of the new Data Collection, whereby the HLT farm avoids duplicate requests to the detector Read-Out System (ROS) by preserving and reusing previously obtained data. Moreover, requests are packed and fetched together to avoid redundant trips to the ROS. Anticipated EB is activated when a large enough portion of the event is requested, reinforcing this effect. A new HLT Processing Unit exploits current architecture trends with a multiprocessing approach that is based on process forking, thereby bypassing thread-safety concerns, while containing total memory usage through the OS's Copy-On-Write feature. HLT and DAQ releases are decoupled by a flexible interface that allows quick updates of the communication between both sides, thus providing increased operational maneuvering. Finally, additional data are recorded through Data Scouting. A method of previewing properties of events whose frequency would otherwise exclude them, this new feature will provide key intelligence for subsequent trigger adjustments.