The breveCluster at NMU uses asynchronous computation among multiple computers for student research into evolutionary learning and artificial life with the breve simulation environment. Current developments for the Fall 2009 semester involve Project Waterhole, wherein multiple trophic cascade scenarios will be simulated, visualized, and analyzed. Past developments on the cluster were brought into being by Correy Kowall and Brian Krent (along with much thanks to efforts Jon Klein for his creation, continual development, and support of the breve software package). A list of various files used by and resulting from the Isthmus project can partially be found under the files folder.

What we've done... Conducted during the Winter and Fall 2006 semesters, Correy Kowall and Brian Krent have completed several evolution experiments of autotrophic reproduction within the breve simulation environment on the cluster. Initial work on this was presented at NMU's Celebration of Student Research & Creative Works and as a department colloquium. The results from these experiments were presented at the international Genetic and Evolutionary Computation Conference (GECCO) held in London in July 2007.

What's current... breveCluster is entering its "third generation"—we humbly started out with salvaged Pentium III systems, which we rapidly outgrew and which were later replaced with HP-Compaq Pentium 4 systems. Now, we've transitioned over to dedicated dual PowerPC G5 worker nodes. Presently, Brian Krent, Joshua Cook, and Matthew Knox (with assistance from Christopher Wells and advised by Randy Appleton) are working on Project Waterhole; the results of which to be presented during the 2009 Argonne Undergraduate Symposium.

Where we're going... Project Latte will aims to replace the breve simulation environment with something a bit more refined—especially in terms of physical simulation by replacing the Open Dynamics Engine with Bullet physics. Further, we plan to combine in our own home-brew AI code, SWIG, and Python to form Latte.)

Status & Info

	Equipment		Info		Notes
	Router & Switch		NetGear, 100 Mbps
	node00 (Head Computer)		PowerPC G5, 1.8 GHz
	node01		Dual PowerPC G5
	node02		Dual PowerPC G5
	node03		Dual PowerPC G5
	node04		Dual PowerPC G5
	node05		Dual PowerPC G5
	node06		Dual PowerPC G5
	node07		Dual PowerPC G5
	node08		Dual PowerPC G5
	node09		Dual PowerPC G5
	node10		Dual PowerPC G5

Log

Date		Notes
October 18, 2009		Doing a test-run of the Project Waterhole codebase. Check our Twitter feed for current updates!
April 22, 2007		Things have been very busy recently, which is why the log hasn't been updated. The breveCluster was moved from NERL to the MA/CS lab awhile back ago, and just yesterday it was moved to its own air-conditioned room. (Thanks to Professor David Powers for getting the air-conditioning installed!) Chris Wells's programming for Project Waterhole has progressed quite nicely. And, Correy Kowall & Brian Krent have submitted their camera-ready copy to GECCO 2007.
February 12, 2007		The adventures of ErgoMan! Correy created a hominid that uses Hebbian learning with feedback to attempt to raise the altitude of the head. See 20070212-ErgoMan3.mpeg and 20070212-ErgoMan5.mpeg. (Clarification note: The ErgoMan code is currently in-development and is not running on the cluster. The cluster is currently running an incarnation of the Isthmus codebase.)
February 11, 2007		node02 has been reinserted into the cluster and is fully operational.
February 9, 2007		A sneak-peek at work in-progress currently being done by Correy Kowall: Body, Rope. (Note to the reader: This is a real-time, interactive, dynamic simulation within the breve environment and not mere predetermined animation.)
January 22, 2007		Optical drive issues discovered with the returned node02, so the entire unit will be swapped out for another one (permanently); that returned unit will go back again for repairs, but we're not going to deal with it ever again, since we just permanently replaced it with something from the pool of other systems.
January 16, 2007		node02 has been returned to us with a new hard disk. (Twas removed from cluster due to drive failure.)
December 7, 2006		The grapher was given a slight facelift, and standard maintenance was performed on all nodes (full system updates across the board).
October 16, 2006		Real-time results grapher is now online, thanks to Alex: brevecluster.nmu.edu/cgi-bin/grapher.cgi
October 1, 2006		Some video footage of recent activity: 20060927-Isthmus-TwoArmedWithCommunication.mpeg, 20061001-Isthmus-TwoArmed.mpeg.
September 25, 2006		Standard maintenance updates applied; all nodes rebooted (as required by updates). Script change in effect on all nodes: breve cycling set to two hours (instead of 20 hours).
September 22, 2006		Lock file bug fixed in A.tz (problem was in the save-results method).
September 15, 2006		Standard maintenance updates applied; all nodes rebooted (as required by updates). Current A.tz running under breve on node00 through node07.
September 14, 2006		Correy's new code is going through a trial run with breve on node01 through node07 and node00 acting as centralized data storage. Initially there were some permissions problems on node00, but they were resolved; and everything seemed to be running fine at the end for the day. Currently using breve version 2.5d1. If problems occur with 2.5d1, then we will download breve 2.4 via subversion and build it ourselves, due to the issues experienced with the current 2.4 pre-compiled binaries available online.
September 12, 2006		Cluster rebuild progress: NFS and breve has been setup on nodes 00 through 07. Still have to do node08.
September 9, 2006		breveCluster "version 2.0" coming soon. All but one of the Pentium III systems were replaced with Pentium 4 systems. Still to do: SMP kernels on new machines, NFS, etc.
September 6, 2006		Doing a clean install of Ubuntu 6.06 on all nodes for the Fall 2006 semester. SMP-686 kernels are currently in use on node00, node04, and node08 to squeeze what little juice we can out of Hyperthreading.
April 2, 2006		Added /Network/Tools to the PATH environment variable on all nodes for our standard user login. Note: File organization will resemble that of the filesystem laid out by GNUstep.
April 1, 2006		Unified breveMaster and breveSlave folders into just /breve to simplify things. Shares (/etc/exports) and BREVE_CLASS_PATH environment variable on head node updated. Share mounts (/etc/fstab) and BREVE_CLASS_PATH environment variable updated on all worker nodes to reflect this change. The local folder /breveSlaveData is now /breveData on all nodes, and C.tz was updated to reflect this change. All old, unused folders removed. Also, there is now a general purpose read-only share on the head node called /Network mounted on all worker nodes.
March 4, 2006		New NFS setup implemented for low-maintance sharing of breve files to the worker nodes.
February 2, 2006		The void which was node04 has finally been replaced with a new HP-compaq system. It's shiny.
January 28, 2006		Since the Windows build of breve seemed to be the offending factor on multiple levels, Linux was installed on the head node. The cluster is now 100% Linux.
January 25, 2006		Debugging breve 2.4 pre-alpha weirdness (still looks like breve itself is to blame; workaround devised and to be implemented tomorrow)
January 23, 2006		Debugging our own code (though, problem appears to be breve's fault)
January 22, 2006		Assembly and shakedown
January 20, 2006		Initial hardware gathered

Documentation, et cetera

Title (and Kind)		Description
Running breveIDE under Ubuntu Linux (PDF) (Text)		Notes on installing/running breveIDE under Ubuntu Linux.
breveCluster Purchasing Wish List (PDF)		Things that would greatly help our efforts.
breve FAQ (PDF)		Minor notes of consequence.
Research Prospectus (PDF)		Nugget of info about research going on in the Fall 2006 semester.
What is breve? (HTML)		Information about the breve environment and the origins of its name.
Notes on screen (HTML)		Scattered notes on usage of the screen utility.