2. EGEE Catalogs Peter Kunszt EGEE Data Management Middleware http://cern.ch/egee-jra1 Service Grids NeSC, 22-23 July 2004 EGEE is a project funded by the European Union under contract IST-2003-508833

3. Service Grids – NeSC, July 22-23, 2004 - 2 High-level strategy for middleware EGEE Middleware –  To re-engineer generic middleware packages  Incorporating experience from EDG, VDT, LCG, AliEn (product from CERN Alice experiment) and others  Architected for scale and performance requirements of LCG and other applications EGEE design team formed early to develop architecture  Architecture: https://edms.cern.ch/document/476451https://edms.cern.ch/document/476451 Fast prototyping approach  Short update cycles to give applications the chance to influence and give feedback

4. Service Grids – NeSC, July 22-23, 2004 - 3 Guiding principles Lightweight (existing) services  Easily and quickly deployable  Re-use as much as possible Interoperability  Allow for multiple implementations Resilience and Fault Tolerance Service oriented approach  Follow WSRF standardization  No mature WSRF implementations exist to date, hence: start with plain WS – WSRF compliance is not an immediate goal  Aim for WS-I compliance Co-existence with deployed infrastructure  Co-existence (and convergence) with existing grid infrastructures (e.g. LCG2) are essential for the EGEE Grid service EDGVDT... LCG EGEE...AliEn

5. Service Grids – NeSC, July 22-23, 2004 - 4 High-level functional decomposition Starting point was the ARDA roadmap document  Focus is upon interfaces that can be composed into useful services

6. Service Grids – NeSC, July 22-23, 2004 - 5 EGEE Data functional interfaces File Catalog  Management of the logical namespace Replica Catalog  Tracking of file replicas Metadata Catalog  Application specific metadata In particular, metadata used to select logical files Combined Catalog  Added functionality by orchestration of the 3 catalogs (providing transaction safety) Storage Element  Where the files get stored  SRM interface (see GGF GSM-WG) Manage a Storage Resource Space reservation Put and retrieve files using various protocols  Posix-like File I/O Most posix-compliant feature support Abstraction over existing MSS IO mechanisms Data Management: File Transfer Service  Reliable transfer of files between two sites File Placement Service  Transfer and register files  Orchestrate File Transfer and Data Catalog services Data Scheduling Service  Event-based data transfer, using File Placement Service

7. Service Grids – NeSC, July 22-23, 2004 - 6 Services: EGEE Catalogs File Catalog  Management of the logical file namespace Replica Catalog  Tracking of file replicas Metadata Catalog  Application specific metadata; in particular, metadata used to select logical files Combined Catalog  Added functionality by orchestration of the 3 catalogs (providing transaction safety)

8. Service Grids – NeSC, July 22-23, 2004 - 7 File Catalog Metadata Catalog Replica Catalog Files and Catalogs LFNGUID Master SURL SURL Metadata

9. Service Grids – NeSC, July 22-23, 2004 - 8 Services: EGEE Catalogs SOA: WS-I Implementation status: prototype

10. Service Grids – NeSC, July 22-23, 2004 - 9 Service Operations File Catalog operations Directory operations Directory permissions Symbolic links List, find +Bulk ops (upload) Replica Catalog operations GUID mappings to SURLs ACLs File ‘stat’ like metadata +Bulk ops (upload, delete) Metadata Catalog operations Query, returning list of LFN/GUID Set metadata based on LFN/GUID Query metadata of LFN/GUID Combined interface List based on LFN (including replicas, metadata) Add entries just based on LFN (auto entry of GUID, SURL) Permissions based on LFNs http://cern.ch/pkunszt/catalogs/

11. Service Grids – NeSC, July 22-23, 2004 - 10 What do you use to build your service? (i.e. How ‘standard’ is your service?) Widely Implemented Standard Specification (1pt)  All services are described through WSDL, WS-I compliant (nontrivial!)  X509 extensions used for authorization (VOMS) Implemented draft Spec (2pt)  GSS/GSI for delegation Implemented draft specification (3pt)  -- Implemented proposal (4pt)  -- Non-implemented proposal (5pt)  -- Concept (6pt)  -- TOTAL: 4 Will use: messaging (JMS) first (+?), WS-Transactions, WS- Notification when they are available with lower rankings Security: Delegation portType (proposal) (+4?)

12. Service Grids – NeSC, July 22-23, 2004 - 11 Service Dependencies What else does your service depend on (i.e. external dependencies)?  RDBMS: need a JNDI connector. Can be anything beneath that, in principle. Implementation currently exists for Oracle, MySQL.  Logging: log4j What does your implementation depend on?  Tomcat 4 or 5  Java 1.4  Axis 1.1  Security libs GSI (using CoG + GSI security libs)

13. Service Grids – NeSC, July 22-23, 2004 - 12 AAA & Security What authentication mechanism do you use?  https: SSL/TLS / TrustManager + CoG What authorisation mechanism do you use?  GSI Delegation -- Working on Delegation portType  VOMS/AuthzManager  Work ongoing on restricted delegation What accounting mechanism do you use?  Logging, RGMA (see Abdeslem’s talk) Does service interaction need to be encrypted?  No. Still waiting on detailed req’s from users whether they really need this If these are not used now, will they be in the future?  Plugin-based extensibility planned. GSI over https used today. Extensions should talk anything that people need (WS-Security in particular)

14. Service Grids – NeSC, July 22-23, 2004 - 13 Exploiting the Service Architecture What features from your ‘plumbing’ do you use in your service?  Factory port : no  Factory pattern : no  Logging : yes  Event notification : not yet  Meta-data : yes  Registry discovery/advertisement : yes  Other OGSI/WSRF/WS/WS-GAF characteristics? No. but interested in –Messaging –Distributed Transactions, Sessions –Notifications and Eventing

15. Service Grids – NeSC, July 22-23, 2004 - 14 Service Activity Multiple interaction or single user?  multiple Throughput (1/per day or 100/per second?)  Many per second. Typical data volume moved in  Lots of small simultaneous single operations  Bulk operations with O(>10000) entries, up to O(10 6 ) Typical data volume moved out  Lots of small single ops (queries for ACL, lookups)  Bulk listings used by scheduler

16. Service Grids – NeSC, July 22-23, 2004 - 15 Service Failure Required Reliability  Support for many bulk operation failure policies Fail on any Try as many as possible [implies Policy management] [implies notification for asynch ops]  Atomic operation failures ‘straightforward’ Required Persistence  State persistence not required, try to be atomic Required Availability  Deployment choice.

17. Service Grids – NeSC, July 22-23, 2004 - 16 Required Service Management Remote access to:  Performance  Progress  Diagnostic and repair interfaces