<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.0 TRANSITIONAL//EN">
<HTML>
<HEAD>
<META HTTP-EQUIV="Content-Type" CONTENT="text/html; CHARSET=UTF-8">
<META NAME="GENERATOR" CONTENT="GtkHTML/4.6.6">
</HEAD>
<BODY>
On Mon, 2013-11-18 at 15:58 +0100, Rick Timmis wrote:<BR>
<BLOCKQUOTE TYPE=CITE>
I have updated the website with the details for the December meeting.<BR>
<BR>
Over to Malcolm for the Beowulf stuff.<BR>
<BR>
</BLOCKQUOTE>
<BR>
Hi Rick et al...<BR>
<BR>
I'm working on getting a complete SD card image with a version of arch linux and the open source mpich software installed and some programs that we can run on it. I've tested it with three nodes, I just need to put some finishing touches to this image which I hope to complete by the weekend so I can make it available well before the next meet (the SD card image will be about 180 mb tar.gz file and is 2GB expanded). <BR>
<BR>
The requirements for the cluster are:<BR>
<BR>
Network Switches (we have 2 x 24 port I believe)<BR>
A number of Trailing extension leads ( I can bring a 2 x 4 port)<BR>
<BR>
As many pies as we can muster! Each Pi needs<BR>
a blank SD Card (at least 2Gb) loaded with cluster image (I have enough for the three pies I can bring and 1 spare)<BR>
a network cable (I can provide at least 5 spares)<BR>
a UK micro usb power supply<BR>
<BR>
For the control Pi station I will bring an screen and keyboard though we could do that via SSH.<BR>
<BR>
It probably makes sense to put a name label on all the items that you bring. <BR>
<BR>
Over on the DCGLUG mailing list there was a discussion about RPI clusters, apologies that I missed it at the time... I use a filter for the DCGLUG messages to keep them out of my inbox, because the discussion was tagged with the DCGLUG subject heading it got filtered and I missed it!<BR>
<BR>
Gordon is correct: I regard this as an academic exercise to learn about parallel computing. We will probably need to have about 20 Pies to make a cluster with the same power as a PC. It would be cool if we could get enough Pies together to beat my laptop.<BR>
<BR>
The Monte Carlo method of Pi calculation lends itself to being done on a cluster since it involves repeatedly doing small independent actions that are aggregated and it's not important which order the actions are aggregated. <BR>
<BR>
With regard to Rick's suggestion of bit coin mining, I'm not sure that is practical on a Pi Cluster and probably not one that is so temporary as ours will be. I'm assuming we can put the cluster together in an hour or so and run it for the rest of the evening then it will be dismantled.<BR>
<BR>
I hope to post more details early next week.<BR>
<BR>
Thanks for your interest,<BR>
<BR>
Malcolm<BR>
<BR>
<BR>
</BODY>
</HTML>