Scalable Distributed Systems

Hobbes: Exascale OS & Runtime

Hobbes is an OS/R framework for extreme-scale systems that support application composition, addresses power/ energy, scheduling and resilience concerns and uses virtualization to provide flexibility for different operating environments.

UNM Investigators: Patrick Bridges and Dorian Arnold
 SNL, LANL, ORNL, Georgia Tech., NCSU, Northwestern, U. of Arizona, U. of Pittsburgh

Scalable Middleware and Tools

These projects aim to improve our understanding about how to design and eploy protocols and infrastructures for HPC systems. Topics include tree-based overlay networks, large scale debugging and scalable job startup.
UNM Investigators:
 Dorian Arnold
Collaborators: LANL, LNL

Indefinite Scalability

Challenging the premise of deterministic program execution, this project aims to overcome architectural scalability limits, like fixed-width addresses and globally unique node names, via the Movable Feast Machine, a robust, indefinitely scalable computer architecture
UNM Investigators: Dave Ackley and Lance Williams

Understanding Power Performance

Can we predict from first principles how energy use scales with system size? Based on biological theories, we provide order-of-magnitude predictions and optimize power performance, as a function of size, and compare with existing chip designs.
UNM Investigators:
Melanie Moses and Stephanie Forrest
Collaborators: UNM ECE, U. of Utah

Autonomic Distributed Systems

This project studies large-scale, loosely coupled, heterogeneous, distributed systems that self-adapt to dynamic changes in resources, workloads, and application needs and goals.
UNM Investigators: Trilce Estrada
Collaborators: U. of Delaware, Northeastern U.