Isn't there a formula that accounts for the # of instances on a node? We have 31 databases in our 4-node non-production 11.2.0.2 cluster and at some point I had to bump up MEMORY_TARGET to 500MB and set PROCESSES to 1000 to avoid memory issues on startup. Best, Jeff On Mon, Oct 8, 2012 at 3:29 PM, rjamya <rjamya@xxxxxxxxx> wrote: > Niall, > > The formula seems to be > > 1) If PROCESSES parameter is explicitly set: The MEMORY_TARGET should > be set to no less than: > 256M + (PROCESSES*132K) -- (64bit) > or > 256M + (PROCESSES*120K) -- (32bit) > > > 2) If PROCESSES parameter is not set: The MEMORY_TARGET should be set > to no less than: > 256M + ((available_cpu_cores * 80 + 40) * 132K) -- (64bit) > or > 256M + ((available_cpu_cores * 80 + 40) * 120K) -- (32bit) > > so if you have 1000 processes set then for 64 bit then your > memory_target becomes (256*1024)+(1000*132) K in size, which is about > 384M, for ASM that is a bit of a jump from 256M.. > > I think it is plus (integer * multipled by 132k) not plus 132k. > > oh well, we tried. > Raj > > On Mon, Oct 8, 2012 at 3:18 PM, Niall Litchfield > <niall.litchfield@xxxxxxxxx> wrote: > > Hi Raj > > > > I don't find that a great formula since it boils down to 256 (mb) plus > > integer plus 132 (kb). Which for all reasonable CPU cores per server is > less > > than 272mb unless there's a missing unit per process. Any time I see two > > different units added together in a 'formula' my hackles rise.... > > > -- > //www.freelists.org/webpage/oracle-l > > > -- //www.freelists.org/webpage/oracle-l