Is this mixed 32/64 test for a 64 build suggesting there's going to be 'in production' performance checking to see which does best on a host? Of course, no doubt you're testing to make sure both results remain equal, just not hoping that quorums are allowed run mixed, when quorum 2 is required.

--//--

There are slight variances in the output for the 32 bit vs 64 bit versions. These have been reviewed and understood by the researchers and do not impact the results. However, this does mean that only 32 bit will be in a quorum and only 64 bit will be in a quorum.

Picked up 5 so far. Started crunching.

Patrick

One part answered, the other part questioned is if we're going to see testing if a 64 bit host does better with a 32 bit science app versus a 64 bit science app [to exact the question]? Meaning if a 64 bit host is found t run the 32 bit faster, the 32 bit app tasks are send and not the 64 bit.

(btw, you bringing 64 bit builds to the table for CFSW suggests the 4x speedup [my Linux quad went from 4.5 hours to 0.95 hours], is not cutting this research from the 1 year mentioned by scientists to < 3 months :O)

--//--

edit: The question relates to this bit quoted from the OP, meaning is this is the case for an individual host, let it continue to run the 32 bit version.

I have noticed that the working set size differs between 32 and 64 bit.
32 bit needs 277Mb while 64 bit uses about 357Mb.