Brokering is an essential element of service-oriented architectures and utility resource sharing.  
Brokers act as intermediaries that match client needs with provider services by aggregating information
about multiple providers and client requests.  In these systems, broker architectures facilitate efficient
and effective service management by distributing services to their "highest and best use".
Brokers must create pairings in a way that accommodates the system's requirements of flexible 
and extensible management to achieve scalability and autonomy.

This project explores broker architectures in the context of two economic, 
service-oriented architectures: Cereus and Tuscany.  These research systems 
are currently under development at Duke and HP Labs, respectively.  In this proposal 
we will explore alternatives for broker architectures, such as the distribution of brokers, 
the mode of client interaction, and the incentives for brokers to participate in and help 
maintain a sustainable system.  One crucial aspect of our broker architecture is that brokers 
organize into a mesh to share system information.  Services and resources flow through this 
mesh according to local broker policies, ultimately resulting in global service management. 
We propose future work to explore the benefits brokers provide to the clients and providers 
in service-oriented systems, as well as investigate how different design decisions in the 
broker architecture affect service and resource allocation outcomes.
Most significantly, we plan to examine how broker allocation policies affect the stability 
and agility of service control and the effectiveness of the broker mesh to match system supply and demand.