• Date: 24 November 2008
• Time: 3:30 PM CET (conversion to other timezones)
• Duration: 1,5h
• Calling number: +41 22 76 77000 (ask to be connected to "GLUE Meeting")
• Shared screen session information:
  • URL: http://glue-wg.glance.net
  • Key: 2411

Agenda

• review and discuss comments from Cal Loomis
  • restart from C.26 on conceptual model comments

Minutes

Participants

• Sergio Andreozzi
• Paul Millar
• Stephen Burke
• Cal Loomis
• Maarten Litmaath
• Laurence Field

C.26

• we agree to have attributes for
  • single job slot
  • mpi job

• WorkingAreaMPIShared
• WorkingAreaMPITotal
• WorkingAreaMPIFree
• WorkingAreaMPILifeTime
• WorkingAreaTotal
• WorkingAreaLifeTime
• WorkingAreaShared
• WorkingAreaFree (only applies if shared working area)

C.27

• Cal: missing information for understanding how to use the values
  • what the model is?
  • to whom is available
  • guarantees
  • lifetime for files
• Maarten: what is the location of the cache?

Action: ask to Balazs to comment on these issues

• clarify the model, do you assume the existence of a certain fixed path or env variable?

C.28 also here to clear the underlying model

C.29 Cal:

• if you have three different disk spaces, then we probably need

• strict definition, clarify the semantics, then define attributes

C.30: MaxTotalWallTime: suggestion to change it into a MaxMPIWallTime, which is the max wall time for an MPI job counted considering a job as a whole (no sum of single slot wall time)

C.31

action: email Balazs and JP, ask for clarification no their requirements on naming we state that second element is middleware name but we have infrastructure/organization name (e.g., nordugrid, teragrid); can they redefine the values using middleware name?

C.32

• for gLite, add
  • org.glite.computingservice,

what about storage?

• e.g. dCache is deployed in different infrastructures
  • org.glite.storageservice no good (not a gLite service)
  • org.glue.storageservice ???

we redefine second component of reverse-DNS naming as namespace; we have default for generic services which is glue;

e.g.: org.glue.computingservice, org.glue.storageservice other options are middleware names (e.g.: glite) or infrastructure names (e.g.: teragrid)

remember that in Service, we have also Capability for abstract functionalities

Action: change definition of Service_t to be namespace-based classification; namespace can be middleware name, organization, or something else; reserve namespace org.glue, org.ogf

• revise values accordingly; add examples for storage services

Action: fix typo in executionmanagement.candidatesetgenerator definition Action: Capability_t values: verify if we can improve naming of values

pending actions

• • C.1: (Sergio) to check how the document should be changed
  • C.5: (Sergio) ask Matt about clarifying his concerns on "specializations" term
  • C.10: (Sergio) will propose a better paragraph
  • C.14 - artf6289 : (Sergio) find extent definition in CIM; add it in the doc; check usage of capacity vs. extent in the document (should be consistent with the semantics)
  • C.15: (Sergio) to put a reference to OGF ref model doc; Sergio to add few words more to explain inheritance
  • C.16 - artf6288 : (Paul) send a draft paragraph for sec section
  • C.17: (Sergio) to sort all data types alphabetically
  • (Sergio) to send new version of conceptual model doc with changes
  • (Sergio): change line from endpoint to activity, square style
  • (Sergio): add motivations to why URI for endpoint ID
  • (James): send examples of endpoint ID for ActiveMQ-based service; link to discussion in W3C about concerns on new URI schemes definitions
  • (Sergio): analyze overlapping between computingactivity and usageRecord; fix terminology
  • (Sergio): Fix C.25 in doc
  • (Sergio): C.26-C.29 send to the list request for comments/examples on the two solutions;
    • what is the relationship between working area, cache, tmp/scratch/appDir, ComputingShare.MaxDiskSpace?