The OS cluster: getting started

The OS cluster: get started

This instruction is based on part of the first homework assignment of CSC591 on cluster computing offered by Dr. Frank Mueller in Spring '03. We also thank Dr. Mueller for sharing this cluster.

Hardware and software configuration

Learn how to compile and execute an MPI program

Log in to an OS lab machine (os20-os35), either on the console or remotely:
```
ssh osXX.csc.ncsu.edu -l <your-unity-username›
```
Your initial password is your student ID.
Please change your password this using
```
smbpasswd -r click
```
From your OS lab machine, log in to the OS cluster (os00-os15):
```
ssh osYY
```
Create the file .rhosts in your home directory and set the access rights as follows:
```
chmod 600 .rhosts
```
Create a subdirectory .gmpi:
```
mkdir .gmpi
```
Then, create the file .gmpi/conf.

Append to your file .bashrc:

export 
PATH="/opt/mpich-1.2.4..8/bin:/opt/gm/bin:/opt/intel/compiler70/ia32/bin:$PATH"
export 
LD_LIBRARY_PATH="/opt/mpich-1.2.4..8/lib:/opt/gm/lib:/opt/intel/compiler70/ia32/lib:$LD_LIBRARY_PATH"

Log out and back in to osYY to activate the new settings.

Find a simple MPI program, such as for calculating Pi linked at the class webpage.
Compile the program:
```
mpicc -g -o pi pi.c
```
Execute the program on 2 processors:
```
mpirun -np 2 pi
```
Try again with a different number of processors.
Add timing operations if necessary and measure the computation time and scalability.

Hints:

MPI API
MPICH
A User's Guide to MPI by Peter Pacheco
Debugging: Gdb does not work with MPI, and we don't have totalview (an MPI-aware debugger). Hence, you are down to good old printf debugging, unfortunately.
Nodes not working: Sometimes, nodes go down and you don't know about it. If your program hangs after mpirun, then check if the nodes are still up:
- ping osYY
- rsh osYY pwd
If ping fails or pwd does not show anything, you may consider the node unusable for now. Remove this node from your ~/.rhosts file and post such problems on the mail list.
I get a message "Unable to get GM local node id".
Issue the command:
```
/opt/gm/sbin/mapper /opt/gm/etc/gm/map_once.args
```

Additional references:

Cliff Notes for GDB - the very bare bones of GDB

Gnu manuals (make, gdb and many more)