Storage Resource Manager Context

1
Storage Resource ManagerContext

• D. Petravick
• NESC
• Oct 4, 2004

2
RunII PB/yr

• Substantial local facilities
• PB/yr for Run II (growing)
• HEP facilities often write tape and pseudo-file
  system infrastructure
• (Subsequent talks)
• Data Movement _at_ FNAL is many 10s of TB/day

3
Evolution in HEP

• FNAL- HEP Run II experiments are increasingly
  exploiting grid resources.
• FNAL CCF Goal
• Push towards operability on the US OSG
• OSG will be useable bu CMS
• Compliments CMS Tier 1 cetenr and US role in CMS
  SC program
• Work Plan
• SRM interface on dcache (will follow)
• SRM interface on sam Cache
• SRM interface in front of Tape
• SRM reference implementation on simple file
  system
• Collaboraiton w/ NeST.
• SRM in ildg (theory)

4
To Do this on the Grid

• What is needed beyond file transfers
• How would you build a cache w/o df?
• How would you command a SS to be a client?
• How can local people know if a file is no longer
  needed?
• How to manipulate namespaces?
• How to manipulate permissions?
• How to do accounting?
• ..

5
Storage Element Model 1
Stage In
Stage In
Stage Out
Grid side
WAN-gtCORE-SS
SSgtworker side
Local Side
srv
srv
srv
srv
srv
(POSIX style I/O)
v
wn
wn
wn
wn
wn
6
Many types of SE

• Diverse SEs motivate acces to SEs via
  protocols, not by reference implementations.
• Temporary SEs
• Have obvious roles in replica systems
• But also roles in creating files.
• Creation of publishable files is complex.
• Variety of hardware
• Physically muxed . High prrf but commodity
• Permanent SEs
• Today Tape is typically involved.
• Brings file size constraints.

7
LHC/OSG Challenge

• CMS Robust Service Challenge
• CERN -gt FNAL
• Mastered by SRM at FNAL.
• Grid FTP _at_CERN
• Managed SE _at_FNAL
• Functionality not performance

8
Most Visionary Future Performance
9
Most Visionary Future Performance
10
Summary

• SRM flows from a SciDAC project
• LBLFNAL collaboration
• Min product is WSDL, independent implementations
• Provide grid access to SS beyond file transfers.
• Enhance the probability that file transfers wll
  succeed, perform.
• Advanced features
• Dynamic/.optical networking Awareness of
  bandwidth between two sites
• Inclusive, not exclusive collaboration