CMU-CS-98-167
Computer Science Department
School of Computer Science, Carnegie Mellon University



CMU-CS-98-167

A Case for Customizable Resource Management in Networks

Peter Steenkiste, Allan Fisher, Hui Zhang

October 1998

This report supercedes CMU-CS-97-195
Darwin: Resource Management for Application-Aware Networks

CMU-CS-98-167.ps
CMU-CS-98-167.pdf


Keywords: Resource management, hierarchical scheduling, active networks, congestion control, virtual networks, application-specific adaptation, network-aware applications.


We envision the deployment of an electronic services market that will deliver a wide range of electronic services over networks. This market will allow applications to combine resources at endpoints with resources inside the network to deliver high-quality products to end-users. Electronic services will range from simple data delivery services to sophisticated value-added services such as video conferencing and data mining. We argue that the deployment of such a diverse set of services will be facilitated by the presence of an integrated set of customizable resource management mechanisms. In this paper, we introduce three concepts that respond to this need. First, the resources allocated to an application will be integrated in a virtual network that forms the basis of runtime resource management and quality of service optimization. Second, each resource is managed by a hierarchical resource manager that satisfies the combined priorities and constraints of the services and applications sharing the resource. Finally, since both network conditions and application requirements can change, application-specific adaptation is needed to optimize quality of service at runtime. The CMU Darwin project is developing a comprehensive set of customizable resource management mechanisms based on these concepts. We outline the Darwin resource management mechanisms and we describe an initial implementation.

21 pages


Return to: SCS Technical Report Collection
School of Computer Science homepage

This page maintained by reports@cs.cmu.edu