|
Cluster Deployment
Possible Deployments
PCM provides the fastest way to build a cluster.
You can mix and match nodes in almost any configuration, so long as each can connect
to any other over a network, even other the Internet!
PCM can do this because it does not mandate a "head node". PCM operates flexibly to fit
your working style, computing resources, and network topology.
PCM is robust because it is capable of handling arbitrary changes in cluster configuration.
No other cluster type is this flexible and capable.
|
 |
|
 Dedicated Cluster
A cluster of Macs dedicated to parallel computing. Can be Xserves, Power Macs, etc.
Such the cluster of 33 Dual-Processor Xserves that achieved 217 Gigaflops
at NASA's Jet Propulsion Laboratory or
128 Dual-Processor Xserve G5s achieving 1.21 Teraflop
at UCLA.
-
+
+
+
Mixed Cluster - Some Dedicated, Some Personal
Some Macs Dedicated for Parallel Computing, while additional Macs normally
used for desktop applications are added during the night or the weekend or other times their users are away.
This approach is in use at the
UCLA Plasma Physics Group.
-
 Macintosh Student Lab
A Mac lab by day, a supercomputing by night. Macs are commonly used in
computer labs at universities. With no hardware modifications, these can be used as a large Mac cluster when the students are away.
This approach was tested at USC,
combining 76 Dual-Processor Power Macs to achieve 233 Gigaflops.
-
+
Impromptu Cluster on the Go
Combine PowerBooks and iBooks and other available machines via combinations of Ethernet and Airport networking.
If you need computational power while on the road, you can even tap into your Mac cluster (assuming there are no firewalls)
back home via the Internet.
When Dauger Research, Inc., gives a presentation, this ability is almost always demonstrated.
Photographic evidence can be seen in
a description of the FOSE 2003 conference, where nodes
in Washington, DC, and California were combined as a computing cluster.
-
+
+
+
+
+
+
+
+
Combine Any of the Above
Any of the above nodes can be combined for parallel computing.
You can be creative!
For example, Professor Timberlake at Berry College
combines his dedicated nodes with 25 iMac G4s in his college's computer lab.
|
