> > But the point of the benchmark results from IBM/DB2 is that the single
node
> > (64 processor machine) seems to scale better than the 16 node (4 CPU per
> > node) Oracle benchmark for TPC-C. Of course to know for sure, the
processors
> > on the Oracle RAC benchmark would have to be upgraded from 1.5 GHz to
1.6
> > GHz (to make Daniel happy) and the number of nodes would need to
increase by
> > at least 2 times to match the IBM/DB2 results.
>
>"Mark Townsend" <markbtownsend_at_comcast.net> wrote in message
> Umm - actually, upgraded from 1.5 to _1.9_ Ghz, as well as 6 Mb to 36 Mb
> L3 caches and 768 GB to 2 TB of RAM. Nbr of CPUs would of course need to
> remain the same (both had 64). Oh yeah, and around $10 million dollars
> of change to spend on something ($6,541,770 cf $16,669,230), perhaps a
> switch or two.
>
The cost/TPM of the DB2 benchmark is slightly lower than the Oracle RAC
cost/TPM. If the Oracle RAC used the IBM 1.9 GHz processor nodes rather than
the Intel nodes, the cost of the Oracle RAC would be much higher than the
DB2 benchmark.
Received on Sat Jan 29 2005 - 23:24:21 CST
