And that should apply to the write-write contention the same way... (and again it's the LOB-segment specific space management logic which dictates the (different) locations where the two writes physically end up writing). Of course if you try to update let's say only a 1kB portion of a 16kB LOB chunk, then the chunk has to be read from disk first. On Fri, Dec 7, 2012 at 1:19 PM, Jonathan Lewis <jonathan@xxxxxxxxxxxxxxxxxx>wrote: > > ----- Original Message ----- > From: "Saibabu Devabhaktuni" <saibabu_d@xxxxxxxxx> > > | Jonathan, > | I couldn't reproduce the scenario of reading the same block with "db file > sequential read" waitevent and immediately followed by "direct path read" > waitevent. > > I thought I did it last night with: > > flashback logging on > lob segment in ASSM tablespace > switch log file > flush buffer cache > insert lob > > -- //www.freelists.org/webpage/oracle-l