Hello Igor, Based on our discussions, I've filed * http://askmonty.org/worklog/Server-RawIdeaBin/index.pl?tid=123 "DS-MRR for clustered PKs: more efficient buffer use" * http://askmonty.org/worklog/Server-RawIdeaBin/index.pl?tid=124 "DS-MRR for clustered PKs: cost function" * http://askmonty.org/worklog/Client-BackLog/index.pl?tid=125 "Make DS-MRR sort the ranges before scanning the index" Could you please check if that correctly describes the conclusions we've arrived at? Also, WL texts contain several unresolved questions, marked with "TODO". BR Sergey -- Sergey Petrunia, Software Developer Monty Program AB, http://askmonty.org Blog: http://s.petrunia.net/blog