EDG Response to HEPCAL Use Cases J Templon

EDG Response to HEPCAL Use Cases J. Templon, NIKHEF/WP 8

HEPCAL History u EDG WP 8 (HEP Applications) started gathering common requirements in early 2002 u RTAG on HEP Common Use Cases was commissioned u WP 8 presented an advanced draft of its use cases document to the RTAG (at the RTAG kickoff meeting) to use as a starting point u RTAG delivered to LCG after approximately four personmonths of work Jeff Templon – ATF Meeting, Budapest, 2002. 09. 03 - 2

EDG approach to HEPCAL Use Case document u Use case document was disseminated within ATF to all work packages as well as EDG Architecture Consultant u Work packages were commissioned to respond whether they believed their product could already support the use cases, or if not did the use case fall within the planned scope of their work package (release 2 or 3) u All middleware work packages (1, 2, 3, 4, & 5) responded u Work package 8 (HEP applications) ATF representative (also a HEPCAL RTAG member) collated the responses and made preliminary replies to the WPs (process ongoing) Jeff Templon – ATF Meeting, Budapest, 2002. 09. 03 - 3

General Comments u Breakdown of comments: number of use cases treated by each WP (43 use cases in total) n WP 1 – 19 use cases n WP 2 – 19 use cases n WP 5 – 8 use cases n WP 4 – 7 use cases n WP 3 – 2 use cases n Architecture Consultant – 33 use cases! u 39 n n use cases were examined by at least one WP. The four exceptions: “VO-wide resource allocation to users”, “Revocation of Grid Authorization”, and “Obtain Grid Authorization” – EDG has no “security WP”, action for security group. “Conditions Publishing” – very HEP-specific, and some aspects of the use case are unresolved (need more input from experiments) Jeff Templon – ATF Meeting, Budapest, 2002. 09. 03 - 4

Global Summary of Responses u Use n n case is already implemented (release 1. 2) – 19 Mostly basic job submission and basic data management For half of these, WP 8 agrees that the functionality is implemented in 1. 2, but the implementation is quite a bit more complex than that outlined in the use case (esp. data management). The release 2. 0 implementations look simpler. u Planned for release 2 – 10 u Will be considered for release 3 – 4 u Use case not detailed enough – 4 n n VO-wide resource allocation to users – HEPCAL did not make strong requirements on security “Job Splitting” and “Production Job” – were purposely vague in HEPCAL due to lack of clear vision of how massive productions will be run on the Grid. One job auto-split into thousands? Or thousands of jobs somehow logically grouped into one production? u Not planned for any release – 7 Jeff Templon – ATF Meeting, Budapest, 2002. 09. 03 - 5

Use Case Classification by EDG Use Cases addressed now (EDG 1. 2) u Basic authorization/authentication u Basic data file use cases u Basic job submission use cases Use Cases addressed in next major release (EDG 2. 0 – end 2002) u Dataset metadata use cases u Advanced dataset use cases u HEP “data analysis” u HEP “data transformation” Use Cases addressed in future releases (2003) u Job metadata (catalog) use cases u Advanced u Resource Job control use cases allocation & reservation Use Cases not addressed through 2003 u All virtual-data use cases u Error recovery for jobs use case u Experiment software publishing Jeff Templon – ATF Meeting, Budapest, 2002. 09. 03 - 6

Meta. Commentary u EDG n n has addressed (or plans to consider) 33 of 43 use cases WPs are (naturally) a bit optimistic, as WP 8 felt the current implementations were substantially more complex than HEPCAL in half the cases Assessment is that release 2 tools will come closer to attaining the HEPCAL-specified simplicity level in these disputed areas u 7 out of 43 use cases are not planned to be addressed in the project n n Virtual data use cases will be addressed by Gri. Phyn. However WP 8 would like to see an analysis of the compatibility between what Gri. Phyn will provide and what the EDG architecture can accept Need for software publishing wasn’t obvious in the beginning. Discussions have started with WP 2, WP 4, WP 6 Jeff Templon – ATF Meeting, Budapest, 2002. 09. 03 - 7

Suggested Next Steps u EDG considers this activity as very important and counts on the availability of the RTAG members to help the EDG architecture group further elaborate the documented use cases u EDG architecture group (with assistance from EDG WP 8 and possibly HEPCAL RTAG) should produce a test-case (making use of EDG APIs) for each use case. Order of development can be driven by classification given above n n Start with use cases addressed by EDG 1. 2 Continue with use cases addressed by EDG 2. 0 then further releases u EDG architecture group is using the use cases to drive the design of its future releases Jeff Templon – ATF Meeting, Budapest, 2002. 09. 03 - 8

A new set of use cases? u Revisiting HEPCAL will be relevant (e. g. with respect to the “production” use cases) when we have significant user experience with Data Challenges. This RTAG could convene in early 2003. u Many comments indicated that we need “system and site administrator” use cases. The HEPCAL RTAG explored how users will interact with the grid. n n n A “Grid Administration Use Case” RTAG is needed to explore the interaction between Site (and other) Administrators and the Grid. This RTAG should wait until EDG has experience with large-scale deployment. This RTAG should convene before the end of the EDG project (or risk losing valuable experience). Jeff Templon – ATF Meeting, Budapest, 2002. 09. 03 - 9