Alessandra, Andrew, Balazs, Steve, Stephan, Greg, Alexey, Julia
Main topic of the discussion - CRR structure
Main changes agreed:
-Definition of the 'processor' in the CRR context
- Adding definition of the wallclock work
-Number of logical CPUs is changed by number of processors
- Adding namespace for the site
- Adding node_type optional attribute suggested by Balasz inspired by Amazon Instance Types: https://aws.amazon.com/ec2/instance-types/
- Adding optional attribute hosting_site in the accesspoints section. In case the accesspoint is located in a site different from the one providing computing resources. If acesspoints are located at the same site which hosts computing resources, this attribute is omitted
Julia told that CRIC could already digest CRR info. The only limitation is that CRIC misses the concept of
computingresource or cluster, which is being currently implemented by Alexey.
Balazs pointed out that it is possible that CRR would require authentication. Julia told that services which would consume this data could authenticate with service certificate, at least the CRIC can do.
Steve shortly described his work on CRR validation.
Validation aims to check that the document is well formed, conforms with the schema and satisfies referential integrity. This work is very useful, it would allow sites to validate their CRR and SRR in the future as well. Julia mentioned that there was already a request for SRR validation. Dimitrios was going to look into it, but most probably did not have time. Julia will point Dimitrios to the work of Steve. Steve will present more about his work at the next meeting.
Julia will create the latest version of CRR following today decisions and will attach it to the twiki page.