On 07/05/2012 14:18, Stephen Burke wrote:
> Testbed Support for GridPP member institutes [mailto:TB-
>> [log in to unmask]] On Behalf Of Alessandra Forti said:
>> Mathematically and in practical terms. The weighted average is not a
>> perfect measure but it is the closest we have, short of mapping the WN
>> to their HS06 value in apel as I proposed to try a number of times. I
>> hope at least the number of job slots (rather than nodes) of each type
>> is close to 50-50 because if you have 1000 of one type and 3000 of
>> another your HS06 hours counting will reflect even less what is
>> happening in reality.
> APEL is a different matter anyway, that comes from the GlueCECapability:CPUScalingReferenceSI00 and it looks like QMUL has 2135 for all of its CEs.
>
> Stephen
I didn't check if he says he is publishing different values for
different CEs I assume that's what he does. The other benchmarks he is
publishing are these
GlueHostBenchmarkSI00: 1770
GlueHostBenchmarkSI00: 2135
GlueHostBenchmarkSI00: 2120
is 2135 correct?
cheers
alessandra
|