Infrastructure Area Infrastructure Area

This area covers middleware services and components that provide common information and management functionality to deployed Grid services. They include the Information System and Service Registry, which provide a view of the services available on the Grid together with their characteristics; the messaging infrastructure, that allows to collect and distribute messages generated by Grid services or user tasks; the service monitoring and management providers that allow the retrieval of Grid services status information and service state management; the Logging and Bookkeeping services that allows collecting, aggregating and archiving job execution information; the accounting functionality to collect, distribute and publish information concerning the usage of resources This area also deals with internal infrastructure components, such as service containers that are required for middleware service.

Information Services

These services enable users, applications and other services to discover which services exist in an infrastructure and further information about their structure and state.

This information conforms to a model which is described using an entity-relationship schema containing denotations of real components of the infrastructure. The information model is an integral part of both the information service and the infrastructure itself.

Messaging

This service is a common component in industrial distributed systems. Recently this technology has been adopted within the EGEE infrastructure as an integration framework for service monitoring and other operational tools.
It has also has been used for publishing event logs from gLite middleware components. So far, Apache ActiveMQ was used for this work. ActiveMQ has also been used by NDGF to change the communications mechanism in dCache.

Accounting

Usage records of resource utilization by user and VOs are required in order to measure the over usage of the infrastructure. Within EMI, two solutions are available, the APEL client and DGAS client. Both generate usage records from the Compute Service and publish them to an external source. Currently no solution for VO-aware storage usage accounting is available at the production level.

Logging and Bookkeeping

This service provides an integrated view of job status, based on information gathered from several Grid components. Recently added features include extended notifications, simple interfaces extended authorization and a non-intrusive purger.

Service Monitoring and Management

Overall service availability monitoring is currently performed via frameworks that implement periodically executed probes. These are mostly in-house solutions like SAM or MonALISA, but also off-the-shelf systems like Nagios. These frameworks are middleware-agnostic and are deployed on top of ARC, gLite and UNICORE. More dynamic Grid service characteristics are very middleware-specific, although these have recently started to converge.

Virtualization and Clouds

There are two broad areas that EMI will address:

  1. The modifications necessary for the compute management services and the information and job definition schemas to accept new requirements to be passed to underlying virtualization managers and request specific platforms or appliances.
  2. The modifications necessary in grid services and cloud management systems to take advantage of the elasticity of the cloud resource management model to make the provisioning of grid services upand down-scale as necessary based on dynamic conditions.

PDF Version


Contact: Laurence Field
E-mail: laurence.field(AT)cern.ch