Executive Director Report 8th July 2008

Key Activities

• Quarterly meetings held with ATLAS, CMS, CDF,D0. This action item list will be used for follow up.

• Reporting LHC Tier-2 site availability and reliability through the WLCG reports. Much progress made in the comparison of OSG and WLCG reporting. WLCG MB has accepted the equivalence of the OSG tests with the EGEE SAM CE and SE tests. OSG Tier-2s expected to be reported fully in the July WLCG reports.

• Certificate Authority acceptance and distribution: Mine is working with the Fermilab security team and lawyers at ORNL and FNAL on the exact requirements of the EAR (Export Administration Regulations). She will be letting the Executive Board know when there is some specific concrete news and action items to discuss. We are considering only distributing CAs when there is a registrant in one or more VOs from that CA. This would drive the process from the VOMS rather than taking all IGTF certified CAs whether any user on OSG is registered with it or not. This would meet the WLCG policy as long as we inform the WLCG Security Officer (Dave Kelsey is the WLCG Security Coordinator, there is no security officer, but we have Ian Birds ok to this process in principle).

Report on D0 usage

• We are receiving weekly reports from Joel Snow of D0 on the efficiency of usage of OSG sites. He measures efficiencies of 10-15% when there is no local SE and <30% when one is present. OSG management is concerned about the low efficiencies and is trying to work with D0 to better understand these. Andrew Baranovski is working on helping sites configure and interface storage elements for D0 use - once the site owners have agree to host this VO opportunistically. We are still a ways from completely understanding the situation. For example, last weeks report from Joel says:
  DZero MC OSG production fell 1.4M events to 2.2M events last week. This is 40% below the average weekly OSG production rate for the past year of 3.6M events.

Production characterization: Bad (with qualification).

A problem with delivery of a certain crucial file needed for a set of jobs caused a significant part of the OSG production loss, and contributed to lower local job efficiencies last week than would otherwise have occurred.

The problem was independent of OSG, having to do with file handling in the SAM system (D0 VO resources) upstream of OSG resources. The problem has been repaired this morning. The magnitude of the impact of the problem is given by the number of batch jobs affected by it = 59*800.

Stale file at UNL dcache required manual removal to unblock data transfer jam.

local job efficiencies: About the same to a little better than last week. For a general OSG cluster I am seeing 6-9% job efficiency. Optimistically speaking, in light of the problem with requests 89862-89870 (VO specific problem) this result is not so bad.

Meetings/Workshops

• Miron, Alain, Alina, Ben Clifford are at the International Grid Summer School in Hungary, which is jointly sponsored by OSG. We have sponsored more than 6 students from the US, some of whom are blogging.

-- RuthPordes - 08 Jul 2008