On 22/06/2015 12:20, Stephen Burke wrote:
> Testbed Support for GridPP member institutes [mailto:TB-
>> [log in to unmask]] On Behalf Of Alessandra Forti said:
>> For me the
>> biggest problem of the BDII is that it has been overloaded with
>> information for slightly different uses (brokering, service discovery,
>> migrations, general description, accounting, passing parameters to batch
>> systems).
> If the information system can be used for multiple purposes I don't see why that's a bad thing - the question should be whether it does in fact work for each purpose, and whether a better alternative exists. Remember that the BDII was originally a temporary hack until we got something better - but despite various initiatives over the years the something better has never arrived.
this multi-purpose is why we have a problem IMO. Several fields are used
for different purposes with a slightly different intention. For example
HS06/KSIsk is there for accounting reasons and have a correct global
value for the CE (there is a constraint on this) and it kind of make
sense to be an average but if used for brokerage it should really by the
minimum value at a site, not to count that we had to add HS06 together
with the fairshares per experiment in a capability field while KSI2k
have their own field. Same goes for memory and queues time limits which
for brokerage should be a minimum but to pass the parameters to the
batch system should be a max. These are two examples, I bet that if
other fields were more used similar problems would come out.
I do find the BDII useful (yesterday I was counting the batch systems
types for example), it is the overlapping information I have a problem with.
cheers
alessandra
--
Respect is a rational process. \\//
|