SAGA and SD OGF 19 Chapel Hill NC

  • Slides: 9
Download presentation
SAGA and SD OGF 19 Chapel Hill, NC Steve Fisher <s. m. fisher@rl. ac.

SAGA and SD OGF 19 Chapel Hill, NC Steve Fisher <s. m. fisher@rl. ac. uk> © 2006 Open Grid Forum

Inputs • Work in g. Lite (EGEE) on SD • Meeting (Copenhagen) • between

Inputs • Work in g. Lite (EGEE) on SD • Meeting (Copenhagen) • between • EGEE, OSG, OMII (Europe and UK), Nordugrid, ARC, Globus and FSU • decided that SD API is a “good thing” © 2006 Open Grid Forum 2

SD architecture in g. Lite Applications/Services Service Discovery API R-GMA Plug-in SQL Query R-GMA

SD architecture in g. Lite Applications/Services Service Discovery API R-GMA Plug-in SQL Query R-GMA Info. Systems © 2006 Open Grid Forum BDII Plug-in XML File Plug-in XPath Query LDAP Query BDII Info. Systems 3 Services publish directly to the information system

GLUE 1. 3 Service • Site may have many services • Services have n:

GLUE 1. 3 Service • Site may have many services • Services have n: n self-relationship • Service may have service data * • (key, value) * * © 2006 Open Grid Forum * 4

SD API • Finding Services • Based on various search criteria • Includes key/value

SD API • Finding Services • Based on various search criteria • Includes key/value pairs (open-ended) • Can use multiple plugins (and combine the results) • Returns a service “object” • Has getter methods • Hide implementation • Allow changes • Optimal efficiency © 2006 Open Grid Forum 5

list. Services List. Services IN Service. Filter. String IN VOFilter. String IN Data. Filter.

list. Services List. Services IN Service. Filter. String IN VOFilter. String IN Data. Filter. String OUT List of service “objects" • Filter strings uses SQL syntax as if it were part of a where clause selecting from a single table. • 3 filter strings • simplifies the implementation, • clarifies the description of the functionality • avoids clash with key name being glue attributes. © 2006 Open Grid Forum 6

Column Names • Column names in the service filter are: • • • Type

Column Names • Column names in the service filter are: • • • Type - type of service Name - name of service Site - name of site Endpoint - will normally be used wth the LIKE operator Service - for associated services • Column names in the VOFilter. String are • VO - will often be used with the IN operator • Column names in the The Data. Filter. String • are taken from the service data key/value pairs. © 2006 Open Grid Forum 7

Examples • list. Services ("Type = 'org. glite. security. voms' ", NULL) • •

Examples • list. Services ("Type = 'org. glite. security. voms' ", NULL) • • List. Services ("Site IN ('INFN-CNAF', 'RAL-LCG 2') ", NULL) • • all services of a matching VO and key/value pairs list. Services ("Type = 'Resource. Broker' ", NULL, "Running. Jobs >=1 AND Running. Jobs <= 5 ") • • all services for matching VOs list. Services (NULL, "VO = 'dteam'", "source = 'RAL-LCG 2' OR destination = 'RAL-LCG 2' ") • • all services matching a type and site name by pattern list. Services (NULL, "VO IN ('cms', 'atlas') ", NULL) • • all services running at any of the two sites list. Services("Type = 'Resource. Broker' AND Site LIKE '%INFN%' ", NULL) • • for a matching type all service matching service type and key/value interval list. Services ("Endpoint LIKE '%Primary. Producer%' ", NULL) • all services matching end point pattern © 2006 Open Grid Forum 8

Going further • Have so far done nothing about conformance to SAGA style •

Going further • Have so far done nothing about conformance to SAGA style • Currently writing a prototype 3 string selection (to replace current large set of calls) • Need to define both • the user API • the plug-in API • Does the approach feel right? © 2006 Open Grid Forum 9