Hi Steve
> Is assessment continuous, now?
Yes, that is the point I was trying to make. The day the current accounting period stops the next period starts. There may at that time be a tweak to change the metrics if they are found wanting in any area, but there is not expected to be a period going forward that does not somehow contribute to future allocations. Maintenance is best not deferred except where the experiments have indicated it is preferred; given that resource contention is expected to increase during this year, deferring may actually be a worse option. What we do want is improved coordination of extended downtimes so that UK delivery of resources is maximised and best aligned with experiment needs. Tier-1s are currently* impacted more by cyclical demands and LHC operations than Tier-2s, though the long shutdown in a few years time may (as an exception) also present opportunities for Tier-2s to undertake major structural refurbishment with decreased impact.
Jeremy
* If the experiments extend out any of the Tier-1 functions to Tier-2s (as LHCb are discussing) this may change for those sites involved.
> (People defer maintenance until a "non-accounting" period, if they can).
> Steve
>
> Jeremy Coles wrote:
>
>> Dear All
>>
>> Last week there was some discussion on the storage list about the GridPP accounting period vs upgrading to recommended middleware, the impression given was that some sites may opt to put off required but disruptive work until the end of the GridPP Tier-2 accounting period. I was confused about this approach because the current accounting period is not tied to periods when higher stability is expected (the need being continuous) and I was not aware that even once the current accounting period ends GridPP would stop collecting metrics for the next period. I checked on this at a PMB last Friday and the expectation is indeed to take the metric data continuously even though for the purposes of allocating tranches of hardware money specific periods have to be used. This is not the same as in GridPP3 where accounting data was taken from specified but discrete periods.
>>
>> This still leaves a question about better coordination of downtimes across sites in order to optimise available resources and maximise availability to users - something that will become increasingly important as we get more contention. This is an area we need to discuss further.
>>
>> Jeremy
>>
> --
> Steve Jones [log in to unmask]
> System Administrator office: 220
> High Energy Physics Division tel (int): 42334
> Oliver Lodge Laboratory tel (ext): +44 (0)151 794 2334
> University of Liverpool http://www.liv.ac.uk/physics/hep/
|