View
0
Download
0
Category
Preview:
Citation preview
EMI is partially funded by the European Commission under Grant Agreement RI-261611
Information System Evolution
EGI Technical Forum 2012
Laurence.Field@cern.ch
EM
I IN
FS
O-R
I-2
61
61
1
Laurence.Field@cern.ch
Before EMI
20/09/2012 2
Service
pull
pull
pull
LCGClients
BDIItop
BDIIsite
BDIIresource
Service
reg
ClassicInfo
Index
ClassicInfo
Server
Service
reg
CIP
UnicoreRegistry
ARCLIB
pull
pull
gLite ARC Unicore
EM
I IN
FS
O-R
I-2
61
61
1
Laurence.Field@cern.ch
Comparison
20/09/2012 3
PossiblyPullNoXMLGLUE 2.0Unicore
Registry
CIPUnicore
YesPullBDIILDIFGLUE 2.0EMIRERISEMI
Pull
Pull
Transport
Model
Not reallyNoLDIFNorduGrid
Schema
Classic Info
Index
Classic Info
Server
ARC
Kind ofTop-level
BDII
LDIFGLUE 1.3GOC DB
(Not EMI)
resource-level
BDII
gLite
FederatedGlobal
Cache
Data
Model
Information
Model
RegistryService
Level
Stack
Focus: Consolidation and Harmonization
EM
I IN
FS
O-R
I-2
61
61
1
Laurence.Field@cern.ch
• Activity started before EMI
– Brought together many stakeholders
– Within an open forum (OGF)
– Produced an agreed standard (recommendation)
• Goal achieved!
– Widely accepted
– 90% of the problem solved
• Agreement on use cases
• Agreement on naming and semantics
– The reset is just a translation and data transport issue!
• Part of the EMI technical plan
– Defined from the start in the DoW
– Requirement from EGI
– Implementation was a major goal for the first year
GLUE 2.0
20/09/2012 4
EM
I IN
FS
O-R
I-2
61
61
1
Laurence.Field@cern.ch
• Service-level information interface
–Fundamental building block
• Recommended Interface
–LDAPv3 interface to GLUE 2.0 information
• We already have 10 years of experience
• It is a requirement from EGI
• Path of least resistance ( low-cost, low-impact )
• Information providers
–Extract information from the underlying service
–Produce GLUE 2.0 information in the LDIF format
EMI Resource Information Service
20/09/2012 5
EM
I IN
FS
O-R
I-2
61
61
1
Laurence.Field@cern.ch
• Syntactical and Semantic checking
–For GLUE 2.0 information
• Can be used as a unit test
–For developers
• Validation test
–For running services
• Will be integrated into ERIS
–Improve information quality
• Ensures only good information is published
GLUE Validator
20/09/2012 6
EM
I IN
FS
O-R
I-2
61
61
1
Laurence.Field@cern.ch
• Service registration and discovery
–A fundamental Grid functionality
• The EMI approach
–Consolidates existing implementations
–Is generic
–Designed for federated infrastructures
–Support policies at each-level
–Incorporated into high-level functions
EMI Registry
20/09/2012 7
EM
I IN
FS
O-R
I-2
61
61
1
Laurence.Field@cern.ch
Consolidated Stack
20/09/2012 8
Service
reg
ERIS
pull
EMIREMI
Cache
Client
pull
pull
pull
EM
I IN
FS
O-R
I-2
61
61
1
Laurence.Field@cern.ch
• ERIS
–Common Service Level Interface
• GLUE 2.0 Information via LDAP v3
–Primary Information Source
• EMIR
–Common Service Registry
• Service Endpoints
–ERIS is an Endpoint
Impact
20/09/2012 9
EM
I IN
FS
O-R
I-2
61
61
1
Laurence.Field@cern.ch
• Top BDII changes
–Use EMIR to discover services
–Use ERIS to obtain information
• No need for site BDIIs
• Pros and Cons
–One less service to manage
–Removes a redundant caching level
–Increases population time
• Issue for dynamic state information
–Query ERIS instead
Future Directions
20/09/2012 10
EM
I IN
FS
O-R
I-2
61
61
1
Laurence.Field@cern.ch
• Push state information
–For power users
• Many sources, many queries
–Use messaging technology
• Already have considerable production experience
–What metrics?
• Need to identify use cases
–What broker topology?
• Are there any other users?
–Monitoring information from services
Future Directions
20/09/2012 11
EM
I IN
FS
O-R
I-2
61
61
1
Laurence.Field@cern.ch
• EMI 2 (Matterhorn) release provides
–ERIS
• Common Service Level Interface
–EMIR
• Common Service Registry
• Top level BDII should use these
–Once they have been rolled out
• Investigate pushing state information
–Using messaging
Summary
20/09/2012 12
Recommended