To move Grid applications from scientific computing to the commercial field, service discovery is a key and challenging issue. Making a service available for consumption by others involves communicating its existence, and providing means to look for it. Therefore, techniques for publication of services (advertisement to a community) have to be established. Discovery defines a process for locating service providers and retrieving service description documents which have been published. Publication and discovery of services can be built upon existing marketplace technologies. Existing related technologies are e.g. the UDDI project or ebXML. Some projects started working on developing a capability for Publication and Discovery, such as the Gridbus project proposes a Grid Market Directory (GMD), which serves as a registry for high-level service publication and discovery in Virtual organizations.

To implement a capability that handles Publication and Discovery the following infrastructural components have to be developed:

• A registry for “Service Marketplaces” has to be created.
• A service provider publishes his service on a “Service Marketplace” along with service prices.
• Creation and registration of resource brokers to discover resources and services with attributes that meet the service consumers QoS requirements.
• Registration of a service consumer to the “Service Marketplace”, who can then either:
  • Inspect the market place himself to find the best service or
  • Assign this task to a resource broker that inspects the marketplace.

These tasks involve several aspects, all covered by existence of such a “marketplace”, which can bring together providers and prospective consumers. This marketplace has to set up procedures for advertising the services (like defining common formats for SLA templates or service descriptions), but also should provide data mining tools to ease the lookup.

Locating Grid services that satisfy users’ requirements effectively and optimally in dynamic and heterogeneous Grid environment is a challenging issue because:

• Services can join and leave Grid dynamically and autonomously.
• Grid services may be persistent or transient. The two types of services need different discovery technologies.
• The status of a Grid service instance may change over its lifetime.