A New Implementation of the Region-of-Interest Strategy for the ATLAS Second Level Trigger Boisvert V Armstrong S Baines J T M Bee C P Biglietti M Bogaerts A Bosman M Brandt S Caron B Casado M P Cataldi G Cavalli D Cervetto M Comune G Corso-Radu A Di Mattia A Díaz-Gómez M Dos Anjos A Drohan J Ellis Nick Elsing M Epp B Etienne F Falciano S Farilla A George S Ghete V M González S Grothe M Kaczmarska A Karr K M Khomich A Konstantinidis N P Krasny W Li W Lowe A Luminari L Ma H Meessen C Mello A G Merino G Morettini P Moyse E Nairz A Negri A Nikitin N V Nisati A Padilla C Parodi F Pérez-Réale V Univ Bern Pinfold J L Pinto P Polesello G Qian Z Rajagopalan S Resconi S Rosati S Scannicchio D A Schiavi C Schörner-Sadenius T Segura E De Seixas J M Shears T G Sivoklokov S Yu Smizanska M Soluk R A Stanescu C Tapprogge Stefan Touchard F Vercesi V Watson A Wengler T Werner P Wheeler S Wickens F J Wiedenmann W Wielers M Zobernig G 2003 Among the many challenges presented by the future ATLAS detector at the LHC are the high data taking rate and volume and the derivation of a rapid trigger decis ion with limited resources. To address this challenge within the ATLAS second le vel trigger system, a Region-of-Interest mechanism has been adopted which dramat ically reduces the relevant fiducial volume necessary to be readout and processe d to small regions guided by the hardware-based first level trigger. Software ha s been developed to allow fast translation between arbitrary geometric regions a nd identifiers of small collections of the event data. This facilitates on-deman d data retrieval and collection building. The system is optimized to minimize th e amount of data transferred and unnecessary building of complex objects. Detail s of the design and implementation are presented along with preliminary performance results. INTNOTEATLASPUBL