Long GC CR MULTIBLOCK REQUEST times

We maintain 10G OBIEE on 2-node RAC, and our nightly ETLs vary significantly in duration.

Now, of course, there are natural variances due to fluctuations in volume of business transaction data from our Ebiz source.

However, these fluctuations also appear to be partly down to extent to which sought datablocks reside in local RAC memory, or far-side RAC.

Furthermore, my experimentations have thrown up what i consider to be a startling finding, that the latencies associated with fetching data via a FTS off disk are consistently less than those associated with getting the same data-blocks across the interconnect (ie: the blocks are all on far-side RAC, so the wait = GC CR MULTIBLOCK REQUEST.