The EGEE Middleware Architecture · 2013. 4. 22. · GridKA School, 23September-2004 - 3...
Transcript of The EGEE Middleware Architecture · 2013. 4. 22. · GridKA School, 23September-2004 - 3...
GridKA School 23 September 2004
The EGEE Middleware Architecture
Leanne GuyEGEE Testing Manager
On behalf of JRA1
www.eu-egee.org
EGEE is a project funded by the European Union under contract INFSO-RI-508833
GridKA School, 23September-2004 - 2
Contents
• The gLite componentsArchitectural designApproach and strategyCurrent implementations
• Integration and testing Infrastructure and status
• From Prototype to ReleaseCurrent gLite status
GridKA School, 23September-2004 - 3
Architecture Guiding Principles• Lightweight (existing) services
Easily and quickly deployableUse existing services where possible asbasis for re-engineering
• InteroperabilityAllow for multiple implementations
• Resilience and Fault Tolerance
• Co-existence with deployed infrastructureRun as an application (e.g. on LCG-2; Grid3)Reduce requirements on site components
• Basically globus and SRMCo-existence (and convergence) with LCG-2 and Grid3 are essential for the EGEE Grid service
• Service oriented approachWSRF still being standardizedNo mature WSRF implementations exist to date, no clear picture about the impact of WSRF hence: start with plain WS
• WSRF compliance is not an immediate goal, but we follow the WSRF evolution• WS-I compliance is important
GridKA School, 23September-2004 - 4
gLite Services
GridKA School, 23September-2004 - 5
WMS
GridKA School, 23September-2004 - 6
CE
• Works in push and pull mode
• Site policy enforcementCEA … Computing Element Acceptance
JC … Job Controller
MON … Monitoring
LRMS … Local Resource Management System
GridKA School, 23September-2004 - 7
Data Management
• Scheduled data transfers (like jobs)
• Reliable file transfer
•SRM based storage
GridKA School, 23September-2004 - 8
Storage Element Interfaces
• SRM interfaceManagement and controlSRM (with possible evolution)
• Posix-like File I/OFile AccessOpen, read, writeNot real posix (like rfio)
SRM interface
rfio dcap chirp aio
Castor dCache NeST Disk
POSIXAPIFile I/O
Control
User
GridKA School, 23September-2004 - 9
Catalogs
Metadata Catalog
LFN
Metadata
• File CatalogFile system-like view on logical file namesKeeps track of sites where data is stored
• Replica CatalogKeeps information at a site
• (Meta Data Catalog)Attributes of files on the logical levelBoundary between generic middleware and application layer
GUIDSite ID
Site IDFile Catalog
Replica Catalog Site A
GUID SURL
SURL
LFN
Replica Catalog Site B
GUID SURL
SURL
LFN
GridKA School, 23September-2004 - 10
Information and Monitoring
• R-GMA forInformation system and system monitoringApplication Monitoring
• No major changes in architecture
But re-engineer and harden the system
• Co-existence and interoperability with other systems is a goal
E.g. MonaLisa
e.g: D0 application monitoring:
MP
P –
Mem
ory Prim
ary Producer
DbS
P–
Database S
econdary Producer
Job wrapper
MPP
DbSP
Job wrapper
MPP
Job wrapper
MPP
GridKA School, 23September-2004 - 11
Security
PEP and PDP will in most cases be part of the secured service, rather than separated services
GridKA School, 23September-2004 - 12
Approach
• Exploit experience and components from existing projects
AliEn, VDT, EDG, LCG, and others• Design team works out architecture
and designArchitecture: https://edms.cern.ch/document/476451Design: https://edms.cern.ch/document/487871/
• Components are initially deployed on a prototype infrastructureSmall scale (CERN & Univ. Wisconsin)Get user feedback on service semantics and interfaces
• After internal integration and testing components are delivered to SA1 and deployed on the pre-production service
EDGVDT . . .
LCG
EGEE
. . .AliEn
GridKA School, 23September-2004 - 13
gLite and LCG-2
LCG-2focus on production, large-scale data handling
• The service for the 2004 data challenges
• Provides experience onoperating and managing a global grid service
• Development programme driven by data challenge experience
Data handlingStrengthening the infrastructureOperation, VO management
• Evolves to LCG-3 as components progressively replaced with new middleware
-- target is to minimise thediscontinuities of migrationto the new generation
• Aim for migration plan by end of year
gLitefocus on analysis
• Developed by EGEE project in collaboration with VDT (US)
• LHC applications and users closely involved in prototyping & development (ARDA project)
• Short development cycles
• Co-existence with LCG-2• Profit as far as possible from LCG-
2 infrastructure, experienceEase deployment – avoid separate hardware
• As far as possible - completed components integrated in LCG-2improved testing, easier displacement of LCG-2
LCG-2 (=EGEE-0)
prototyping
prototyping
product
20042004
20052005
LCG-3 EGEE-1
product
les robertson - cern-it-13
LCG
GridKA School, 23September-2004 - 14
From Prototype to Release
• Prototype setup at 2 sites (CERN & Wisconsin)~45 users registered2nd release mid August
• Many bug fixes• New functionalities
More nodes (more powerful) being added at CERNBeing ported to SLC3 (by end of September)
Second VO being set up (core services at Wisconsin)• Will use globus RLS
GridKA School, 23September-2004 - 15
Integration
• Continuous integration system is in placeCurrently running on RH Linux Enterprise 3.0134 gLite modules built continuously every 60 minutesAutomatic build failure notifications to developersAutomatic packaging of modules in three different formats (source tarballs, binary tarballs, RPMs)Complete nightly builds run every night at 03:30, all packaged deployed to the gLite web site. All builds are tagged for full reproducibilityCoding guidelines, unit test and coverage automatically run for all packages (Java only, C/C++/Perl on the way)
• SCM convergence being reachedSCM plan: https://edms.cern.ch/document/446241
• Common service configuration being worked out• Deployment scenarios being worked out
GridKA School, 23September-2004 - 16
Build System
GridKA School, 23September-2004 - 17
Testing
• Testing sites (RAL, NIKHEF, CERN) up and running24 Machines distributed across the three sites and growingMigrating to RHEL at all sites nowQuattor based automatic installation of testbed beginning at all sites
• Active testing of the prototype Automatic installation testing of rpms/tarballs from nightly buildActive testing of the gLite prototype
• 91 bugs submitted by test team to date• gLite IO, R-GMA and WMS currently primary focus
Testing effort focused on components for pre-production service• Test plan released
Based upon architecture document, release plan, and NA4 requirementshttps://edms.cern.ch/document/477697/
GridKA School, 23September-2004 - 18
Installation testing reports
GridKA School, 23September-2004 - 19
Deployment considerations
• Interoperability and co-existenceExploit different service implementations
• E.g. Castor and dCache SRM implementationsRequire minimal support from development environment
• Sites required to run globus and SRM (might not be required for tactical storage)
Flexible service deployment• Multiple services running on the same physical machine (if possible)
• Platform supportGoal is to have portable middlewareBuilding & Integration on RHEL 3 and windowsInitial testing (at least 3 sites) using different Linux flavors (including free distributions)
• Service autonomyUser may talk to services directly or through other services (like access service)
• Open source software licenseBased on EDG license
GridKA School, 23September-2004 - 20
Current Implementations
• WMSAliEn TaskQueueEDG WMS (plus new TaskQueue and Information Supermarket)EDG L&B
• CEGlobus GatekeeperCondorG (CondorC)“Pull component”
• AliEn CE• EGEE MON
GridKA School, 23September-2004 - 21
Current Implementations Cont’d
• SEExternal SRM implementations
• dCache, Castor, …• LCG disk pool manager (tbi)
AliEn aio (re-factored to gLite-I/O)(
• CatalogsAliEn FileCatalogRLS (globus and EDG)Combined Catalog Interface (tbi)
• Data SchedulingStorkPhedex (concepts)Data mgmt interface and VO data scheduler (tbi)
• Data TransferGridFTP
• Metadata CatalogSimple interface definedAssumption that it will be managed by applications
• Information & MonitoringR-GMAFor application monitoring the aim is to plug in other systems as well.
GridKA School, 23September-2004 - 22
Currents Implementations Cont’d
• SecurityVOMS as Attribute Authority and VO mgmtmyProxy as proxy storegridmapfile and GSI security as enforcement
• Plan is to move to more fine-grained authorization (e.g. ACLs)
• Plans with globus to provide a set-uid service on CE
• AccountingEDG HLR
• User InterfaceAliEn shellCLIs and APIs
• Move to autogenerated APIs from WSDL
GAS• New development
• Package managerExplore existing solutions
GridKA School, 23September-2004 - 23
Summary
• Next generation middleware being designed and assembledPrototype first tangible outcome
• BUT this is a PROTOTYPE not a release!Architectural and design work well advanced
• Architecture document sent to EU– Subject to changes
• Design document (draft) exists: https://edms.cern.ch/document/487871/Incremental changes to prototype
• Feedback from applications and operations essential!• Detailed release plan worked out
https://edms.cern.ch/document/468699• First components for pre-production service during autumn• Continuous integration and testing scheme defined and adopted• Technology Risk
Will WS allow for all upcoming requirements? Divergence from standards
GridKA School, 23September-2004 - 24
Links
• JRA1 homepagehttp://egee-jra1.web.cern.ch/egee-jra1/
• Architecture documenthttps://edms.cern.ch/document/476451/
• Release planhttps://edms.cern.ch/document/468699
• Prototype installationhttp://egee-jra1.web.cern.ch/egee-jra1/Prototype/testbed.htm
• Test planhttps://edms.cern.ch/document/473264/
• Design documenthttps://edms.cern.ch/document/487871/
GridKA School, 23September-2004 - 25
gLite Web site