JClarens

1
JClarens

• A Java Framework for Grid Services
• Michael Thomas ICWS 2005

2
Outline

• Background
• Clarens
• JClarens
• JClarens Services
• Future Work

3
CERN LHC
Alice
CMS
ATLAS
CERN Large Hadron Collider
LHCb
4
CMS
The collection of Tier-2 centers is the
primary resource that individual physicists at
universities will depend upon for physics data
analysis.
5
OSG

• Building a sustainable national grid
  infrastructure for science in the U.S.

6
Outline

• Background
• Clarens
• JClarens
• JClarens Services
• Future Work

7
Clarens

• Clarens is a Web Service framework for developing
  and deploying web services for CMS and the OSG.
• Provides a common platform for middleware
  developers.
• Two implementations Python and Java

8
Features

• Javascript-based browser UI
• Support for multiple message-level encodings
• SOAP, XMLRPC, JSON, (RMI)
• Common core services
• Authentication
• Authorization
• Group/User management
• Proxy Certificate management
• File access
• Ability to plug in new services

9
Outline

• Background
• Clarens
• JClarens
• JClarens Services
• Future Work

10
JClarens for developers (I)

• Java implementation of the Clarens Web Service
  framework
• Based on Tomcat, Axis, and other OSS projects
• RPM installer for core JClarens and add-on
  services
• Plug-in new implementations of existing
  services
• service.ltnamegt.implorg.foobar.myClass
• Drop-in implementations of new services
• service.ltnamegt.propertiesltnamegt_service.propertie
  s

11
JClarens for developers (II)

• Build tools for adding new services
• Internal service APIs to assist in writing higher
  level services
• Proxy certificate for invoking other web services
• Shell API for running command line programs as a
  local user (more on this soon)
• Configuration API for reading external
  configuration files

12
HelloWorld service

• Template for adding new services
• Ant buildfile
• build.properties, build.xml
• wsdl2java, java2wsdl, classes, install,
  rpm, binaryzip
• Service-specific configuration
• ltnamegt_service.properties
• Service API
• ltnamegt.wsdl
• RPM spec file
• jclarens_ltnamegt.spec

13
Outline

• Background
• Clarens
• JClarens
• JClarens Services
• Future Work

14
Discovery Service

• Dynamic Web Service registry
• Publish and Query for Web Services
• Uses MonALISA a Jini-based monitoring framework
• UDDI bridge exists to read/write from UDDI
  registries

15
Sphinx Grid Scheduler

• Directed Acyclic Graph scheduling across grid
  sites
• Subscribes to real-time monitoring information
  from MonALISA
• Modular
• Plug in new scheduling algorithms
• Database-backend for robustness

16
Sphinx Grid Scheduler
17
Data Location Service
Discovery Service
MonALISA

• Optimized dataset location for CMS
• Fast accessibility
• Lowest resource cost
• Leverages existing CMS Dataset Catalog services
  and realtime-monitoring information

JClarens Host A
Data Location Service
Dataset Catalog Service
Replica Optimization Service
MonALISA
JClarens Host B
Data Location Service
Dataset Catalog Service
Replica Optimization Service
18
Shell Service

• Securely execute command line shell programs
• Map x509 certificates to local users via
  grid-mapfile
• Setuid program switches to mapped user before
  execution
• Programs run in a sandbox
• Input/output files can be read/written using file
  service
• Internal API can be used by new services
• Map certificates to local users
• Both blocking and non-blocking execution of
  commands

19
Outline

• Background
• Clarens
• JClarens
• JClarens Services
• Future Work

20
Future Work

• Java-based webstart interface
• Shell service to optionally use OSG GUMS service
  for local user mappings
• Generic Discovery Service, not just for web
  services
• Better access to internal database for add-on
  services
• Job Monitoring Service
• Estimators
• Steering service
• Single command for creating new service workspace
• Add support for unit tests to new service template