Our research in formation modeling and simulation depends on the
high-speed, distributed computation resources for research validation,
experimentation and design space exploration, and preparation of
and products for infusion into other programs. To this end we have
several low-cost supercomputers, modelled on the Beowulf-class
supercomputers pioneered by Donald Becker at NASA's
Goddard Space Flight Center.
The first of these was a 12-CPU cluster computer constructed from
independent off-the-shelf PC computers. Connected by an ordinary
Ethernet switched network, this proof-of-concept computer was used
various communication, synchronization, and other attributes of
computing critical to high-speed simulation.
Viewfinder (seen at right, bottom of rack) is one of our current
computer testbeds. Constructed of five dual-cpu machines, it contains
Athlon MP 1800+ processors, 10 gigbytes of RAM, and 0.8 Terabytes
storage. For connectivity, it includes a dual-bonded, switched, fast-ethernet
backbone, and additionally has a 1.3 gigabit, low-latency SCI backplane
well. Our 5-spacecraft formation simulator reaches maximum performance
this cluster, utilizing all ten CPUs.
Our innovative ShuttlePC cluster (seen at right, top of rack) consists
single-cpu Shuttle XPC SS-40G small form-factor (SFF) computers,
an Athlon 1800+ CPU, 512 Megabytes of RAM, and not much else. These
nodes are connected via switched fast ethernet to a Shuttle XPC
that serves as a boot host. This diskless configuration allows
us to upgrade
the cluster software and OS by simply rebooting the system after
master node. A spare master node makes it easy to split this cluster
portable testbeds for demonstration purposes.