Upload
frederica-campbell
View
217
Download
0
Embed Size (px)
Citation preview
Owen Synge Title of Talk Slide 1
Storage Management
• Owen Synge– Developer, Packager, and first line support to System
Administrators.
• Talks Scope– GridPP for the year ahead.
• Role– RAL employee, RAL data management Team.– Working within GridPP2.– Mass Storage and Local Data Management.
• Project Members– Owen Synge, Jens Jensen, Tara Shah, Glen Johnson
Owen Synge Title of Talk Slide 2
Summary
• Background• History• Current State• Mass Storage
– Mass Storage Future– Mass Storage Future : Upcoming Releases– Mass Storage Future : SRM Release– Mass Storage Future : SRM News
• Worker Node Disks– Local Storage Future : Time Scales– Local Storage Future : Priorities/Features
• Storage Future• Conclusion• Issues
Owen Synge Title of Talk Slide 3
Background
• ADS Service– Peta-byte level Tape Storage Solution
• Grid Summary– Distributed Computing– Commodity Hardware
• European Data Grid Project (EDG)– Europe wide Grid Research Project– Provided much of LCG infrastructure
• SRM Collaboration – Collaboration between Fermilab, Jefferson Lab, Lawrence
Berkeley, RAL, CERN– Contributed to the design of the SRM version 2 protocol
Owen Synge Title of Talk Slide 4
History
• EDG SE History– V1
• HTTP based Storage interface
– V2• Java Web services Storage Interface
– V3 (Not History as Still on V2.2)• SRM Standard Compliant Interface
Owen Synge Title of Talk Slide 5
Current State
• EDG SE Status– Stable Version
• Stable since November.– One Bug Fix (File Truncation)– One Administration Script created.– Atlas Data Transfer
» About 1500 files; 2 TB in total transferred to CERN• Going into LCG Production (Security bug on Client Side).
– ADS Support– Not using Disk, Castor or HPSS interfaces
– Current Development Version • Released this month.• Configuration Upgrade (Layered Template System).• Metadata Upgrade (See Later).
Owen Synge Title of Talk Slide 6
Mass Storage Future
• Addressing issues in EDG-SE– Metadata/Configuration (Due in near future)– Scalability (Number of Files)– Performance (Time taken and resources used)– SRM Compliance
• Addressing issues in GridPP– Disk Resources as MSS system– Checksums (Atlas request)
• Addressing issues for LCG (UK)– Clustering for Bandwidth– Testing frameworks
• Addressing issues (Generic)– Access Control– Job/User Namespaces
Owen Synge Title of Talk Slide 7
Mass Storage Future : Upcoming Release Time Line
• Release 2.2– Renaming of LCG release
• Release 2.3 (May/June 2004)– Layered Configuration– Metadata upgrade
• Release 2.4 (Before August 2004)– LCG Release of 2.3 when Stable and bugs
squashed• Release 3.0 (Before August 2004)
– SRM compliance• Release 3.1
– To be decided
Owen Synge Title of Talk Slide 8
Mass Storage Future : SRM Release
• Extra Features– Multiple Files acted upon with a single operation– Fully Asynchronous (srmStatusOfGetRequest)
• Benefits– Interoperability with major Storage solutions in
Grid community.– GFAL and a large number of other Client tools
available.
Owen Synge Title of Talk Slide 9
Mass Storage Future : SRM News
• SRM2– Finalised now working on 2.1 Revisions
• SRM and the GGF– SRM is going to be a GGF standard (Honolulu)– Specification of Basic/Advanced SRM
• We will provide a service somewhere between these specifications for ADS on the first SRM release
• SRM and Other Storage solutions– SRB
• On going work to support SRM API
– DCache • Currently supports SRM v1 API
Owen Synge Title of Talk Slide 10
Worker Node Disks
• EDG/EGEE Local Data Management– Missing from EDG
• Clean Down Worker Nodes• Reservation
– Remote File Access• RFIO, SlashGRID, Replica Manager/EDG SE Client tools,
GFAL.
• DCache– Aggregate Worker node space into storage– Mature system
Owen Synge Title of Talk Slide 11
Local Storage Future : Time Scales
• PM03 LD1: UK Site Coordinated Deployment and Support Plan for Local Storage Management (LSM) system.
• PM05 LD2: Release of LSM integrating with LCG3 at UK sites. LCG 3 expected in PM05.
• PM07 SD3, LD3: Software prototype release integrating with EGEE DJRA1.3
• PM16 SD4, LD4: software prototype release integrating with EGEE DJRA1.6
• PM26 SD5, LD5: Release of MS and LSM integrating with EGEE follow-on.
Owen Synge Title of Talk Slide 12
Local Storage Future : Priorities/Features
• Requirements– Disk Clean down?– Using Worker Node disk space?– Space management?
• Specification– Need to establish which requirements are
highest priority.
• Implementation– Need to get the plan signed off.– Need to employ a new member of the team.
Owen Synge Title of Talk Slide 13
Storage Future
• SRM (De facto Grid Standard)– SRM1 moving to SRM2– GSM Basic moving to GSM Advanced
• Worker Nodes– Clean down after Jobs– DCache
• Must not forget Tier 2+3 have more storage space than Tier 0+1
Owen Synge Title of Talk Slide 14
Conclusion
• SRM– On going evolution of a standards management
API– Not yet clear where access control is going exist
in this area.– No SRM2/GSM Advanced implementations yet
exist
• LSM– Local Storage management solutions not yet
clear
Owen Synge Title of Talk Slide 15
Issues
• Tracking SRM scope and model • Local Storage Priorities• Recruiting and Training new member of the
team• Testing environments for LCG/EGEE• Representation of files (Metadata Group?)
– Trees/UID– Namespaces
• Job/Service/User/VO/host based