While not as comprehensive as Jonathan's experiments mentioned below, in general the "overhead" for RAC binaries can be easily viewed in the Enterprise Manager "Top Activities" form. Waits on event class "Cluster" are colored (coloured) gray (grey).
On 5/8/2014 8:32 AM, Jonathan Lewis wrote:
-- "Also, quite a few years ago now, I believe James Morle did some testing of single-instance on RAC vs. non-RAC binaries, and found a measurable overhead in using RAC binaries for a non-RAC database."That might have been me, rather that James.I certainly did some experiments that showed something like 10% difference in cache management between linking with and without RAC, but in the end I think it turned out to be a bug - one of the developers had left some debug code in the RAC lock-element code path that had shipped. I don't recall whether there was any significant difference in the next version up. But it was in 2007, so I don't think anyone will still be running that version.Regardslok Jonathan Lewis http://jonathanlewis.wordpress.com @jloracle
