The participants are expected to arrive by Monday early afternoon. The official program starts with a Welcome session at 16:00, followed by a dinner (around 19). The conference room is at our disposal 24 hours a day from Monday 16:00 till Friday 12:00 and we can use it for late/early coding sessions :). Breakfast, lunch, dinner is served at the hotel and there'll be also a morning and an afternoon cofee break (all included in the fee). The coding week ends with a Friday lunch.
The technical topics we plan to cover are listed bellow. The actual schedulling of the sessions will be done on Monday and is expected to be changed dynamically during the week. Nevertheless, a preliminary timetable is given bellow.
Preliminary timetable:
[Monday Welcome] 17:00 - 19:00
-logistics
-"how we are going to work together",
-goal of the week, quick overview of topics,
-agenda update including eventual parallel sessions
-Documentation
[Tuesday morning]
- Server-side logging review
- Profiling, defining performance data to be collected
- ATLAS data integration
[Tueday afternoon]
- Performance metrics
- Consistent VO handling
[Wednesday morning]
-ARC.CONF 1
[Wedneday afternoon]
-Simplifications
-ARC.CONF 2
-JURA
[Thursday morning]
-Controldir
-arc.conf (gridmanager, data, gridftp)
[Thursday afternoon]
-arc.conf (infosys, cluster, queue)
-various backend bugs
[Friday morning]
-howto collect performance metrics
-the road to arc.conf changes
-summary and follow up
-
ARC.CONF
Everything related to server-side configuration. Restructuring, cleanup, inconsistency resolution, default values. Pilot-friendly streamlined config.
-
Profiling and logging
Data collection to understand possible ARC bottlenecks. The content and the infrastructure to collect performance data from the various ARC server-side subsystems. Rationalization of the server-logs.
-
Consistent handling of VOs
Everything related to how VOs are treated by ARC CEs: info publishing, authorization, accounting/logging, identity mapping. The client-side SDK should be updated/synched too.
-
Accounting subsystem
Manageability features. Conceptual clarifications (cpu/walltime, multi cores/nodes, ur-logger feature completeness)
-
Controldir reengineering
Document how various processes, components use files in the controldir. Propose more efficient access pattern, restructure files and their file content based on the way those are used. Design of the new "common batch layer".
-
Simplifications
Try to go for the low-hanging fruits. Remove obsolete interfaces, functionality. Infosys jungle cleanup, Job management interfaces. Towards the future centralized WLCG infosys vision. Client SDK.
-
Documentation
Identify the key documents, estimate their freshness. Web/Wiki status.
