Intel® oneAPI HPC Toolkit
Get help with building, analyzing, optimizing, and scaling high-performance computing (HPC) applications.
2059 Discussions

Initial Cluster Install Help Needed


Hopefully this is in the right place. I recently acquired 5 1U rackmount server nodes from a business and would like to cluster them for my own private use for MPI programming. (Research for my degree mostly) 

I have not successfully set up a cluster before, and am having issues figuring out where to start. I've looked at many older guides on clustering, but many leave out important setup details or utilize software that is long retired. I was wondering if anybody had any suggested material/advice for getting this mini-cluster set up.

My hardware includes:

5x identical 1U server nodes: 2x Intel Xeon Dual Core procs, 2GB main memory, 80GB HDD (except master node which has a 320gb HDD), two  ethernet ports on board (one 10/100 and one gigabit)
1x old Intel Centrino laptop
2 Ethernet switches (5 port gigabit and 8 port 10/100)

I should also note: these are older mPGA604 socket Xeons. They are not newer computers!

The home network is simply a central router (Linksys E3000) that serves internet access to my laptop and my roommate's laptop wirelessly. The LAN ports are unused. I don't particularly wish for the cluster to be connected to the internet unless I am doing software updates (security reasons given I think I would like SSH access between nodes). I'll probably end up connecting an ethernet cable to the master when I need internet connectivity to do updates.

I was hoping to use the newer Ubuntu distribution where possible (12.04 specifically, 12.10 I haven't played with yet!). Ubuntu Server 12.04 seemed to be the ideal candidate but I can't figure out how to get all the packages installed (apt-get cannot complete due to no internet connection on the nodes)

I was also considering possible setup options:
~I'd like to be able to push updates and such to the nodes from a single machine. No logging in and installing updates on each machine individually. 
~Ideally I'd like to be able to remote into the master node from my main laptop (from within my home network) and run the programs that way, maybe using WinSCP/PuTTy. This is how we interact with the university cluster, so I'd like to do something similar. I don't want to have to log in and develop on the master node itself
~It would also be nice to be able to control the cluster's power on/off functions remotely from my machine if that is possible. i.e when I power down the master, the compute nodes go with it. Would it be possible to do this for shutdown and startup? I have seen a couple posts about it, but they were very vague as to whether they were actually doing what I want to do.
~Of course, I'd like the cluster to be scalable. I plan to add GPU nodes at some point for Massively Parallel development 

 Apologies for the walls of text, I'm just trying to introduce as much pertinent information as may be needed so that I can get going down the right path. I am rather anxious to get this cluster up and running as I'd love to get started on home-brewed MPI development. Plus, what tech nerd wouldn't want their own cluster? :D

Thank you in advance, and again I hope I put this in the right forum!

0 Kudos
3 Replies
I would recommend the following Intel Cluster Ready recipe using non-commercial software.
0 Kudos
I will read through that reference! Will it be a problem that my machines are older mPGA604 Xeons and not Sandy Bridge chips?
0 Kudos
Black Belt
I don't know how to infer which CPU you have on the older Xeon motherboards, but that shouldn't be old enough to be a problem. More of a problem may be the attempt to include an old laptop on the cluster. At the very least, you need to be able to run the same linux version (one which is covered by the recipe) throughout.
0 Kudos