[Beowulf] A Good Linux Distribution to Start with?

erwan erwan at mandrakesoft.com
Sat Sep 18 05:04:13 PDT 2004


Hi Robert,

>Mandrake ditto -- it has its (IIRC) "CLIC" cluster-specific packaging.
>In both of these latter two cases there are again issues of licensing
>and charges on a per node or per cluster basis.
>  
>
This is not true, CLIC is a researched based product and there is no 
licensing issue inside.
CLIC is fully GPL or BSD like license. Mandrakesoft had started a fork 
called Mandrakeclustering which is "per set of node" licenced based.
As we sold it, the MAUI license apply to us and we must give some fees 
to supercluster.org.  What I call a "per set of node" is we provide a 
product for a maximum number of node and not for each nodes.

Our main goal is to provide people clusters ready to run for people who 
want to use their clusters instead of doing some sysadmins tricks all 
the time.

I'm leaving the company but it seems that a new release of CLIC could 
happen soon.
[...]

>...my major beef with most of the cluster distributions is that they
>really require one to bend the simplicity and scalability and
>customizability of repository-based, package-based installation and
>maintenance schema.
>  
>
That's our choice in CLIC/Mandrakeclustering \o/

>In my opinion, the "best" way to install a cluster is from a repository
>via PXE and something like kickstart if not kickstart itself, where the
>only thing that differentiates a cluster node from a desktop workstation
>is the selection of packages installed and some post-install
>configuration.  An acceptable variant of this would be the newer
>diskless cluster approaches, provided that the exported/cloned node
>image is package-level controllable and can be kept up to date relative
>to a well-maintained mirror tree of repositories with a tool like yum or
>apt.
>  
>
I agree with you robert. I take the opportunity to explain our point of 
view.
First you install the server of the cluster using the cdrom. Then by 
answering 10 questions the server is autoconfigured (DHCP, PXE,TFTP, 
DNS, NIS, AUTOFS, WWW, SSH, URPMI (a rpm installer), etc..)
Once its done, you install the first using PXE (you can make a manual 
installation using DrakX (the GUI) or in automatic (installation only 
ask for the keyboard type and the root password).
Once the node is rebooted, you install what you want on it and tune it 
like you want.
Then you duplicate it using the KATools, It just needs you to power-up 
the new nodes (PXE boot). Nothing else :o)
Everything will be done for you, a node name will be assigned, the linux 
configuration (net & scsi modules) will be automaticaly done.
3mn after the power up (for a set of 250 nodes max, the golden node 
~1GB) all the nodes are ready to compute.

>[...]
>  
>
>RPM isn't perfect, but it isn't bad and it is in wide use and has smart
>people actively working on it to improve it further.  
>
We'd improved rpm via urpmi. We are able to make some // rpm commands 
via urpmi using the KA tools (the best efficient copy method is insured)
The follwing syntax show how to install the povray-mpich rpm and its 
dependencies on the set of nodes called cluster.
urpmi --parallel cluster povray-mpich

AFAIK, apt & rpm are not able to do that :o)

Our goal is also to provide the best tools in the cluster field even it 
is redundant. Everyone could use the tools he loves. For example 
wulfstat is included & configured in Mandrakeclustering 2 and maybe CLIC 
3 (is its released).

Hope it helps too,



More information about the Beowulf mailing list