We are investigating network protocols that support a contracting system for providing various levels of guaranteed quality of service (QoS) for EOSDIS distributed scientific applications. These QoSs are in the form of minimum throughput, maximum delay, and maximum delay jitter, per requested channel. A decentralized pricing policy determines the cost of levels of service based on supply and demand of available resources. In a large wide-area network with resources owned by different organizations, and which must be able to grow in time, a decentralized pricing system is essential to achieve scalability.
Our work to date has focused on a new network switch service discipline, called Leave-in-Time, that provides end-to-end performance guarantees for throughput, delay, and delay jitter. Leave-in-Time determines required memory buffer sizes and link bandwidths for the performance bounds it provides. Pricing is then based on the use of these resources.