Try to log on to one of the three RS/6000 machines in the Computer Lab 2105 NW. They are:

lawrencium.chemistry.ohio-state.edu		(lr.chemistry.ohio-state.edu)		right
nobelium.chemistry.ohio-state.edu		(no.chemistry.ohio-state.edu)		middle
mendelevium.chemistry.ohio-state.edu		(md.chemistry.ohio-state.edu)		left

You don't have to physically sit in front of these computers to submit a Gaussian job; you can logon from anywhere in the Chemistry Department. Your chemistry username and password will work, and you will find yourself in the same /home directory as on chemistry.

Each of the three machines (as well as the chemistry server) has a large amount of disk space in /scr (about 4 GB) which is shared among all users of these machines. There is no disk quota on /scr (unlike on your /home directory). Each of the machines has its own /scr; you can access another machine's /scr filesystem as

      /net/computername/scr

e.g.

      /net/mendelevium/scr
      /net/chemistry/scr

To summarize: /scr has a lot of space (no quota), you share that space with other users, and each machine has its own /scr. Your /home directory has comparably much less space, it belongs to you alone, and you are placed in it when you logon to any machine.

You may choose to store all your input and output files in your /home directory. If you do many calculations, you may find it more useful to store everything on /scr for the time being. But keep in mind that /scr is not meant for permanent storage; it is not backed up, and you should move files that you want to keep indefinitely to your /home directory. If you want to use /scr, our convention is to use subdirectories of /scr/tmp for Gaussian calculations:

      cd /scr/tmp

Create a directory there and name it after your username:

      mkdir username

Then change directory to that location:

      cd username

To run Gaussian, call your input file a name with .com as the extension (such as water.com). To run the job, use the command:

      subgauss queue_name job_name

where for a water.com job to run on the lr_1 queue, it might be:

      subgauss 1 water       (no need to say .com)

Each of the 3 machines has 3 queues for running computational jobs. They are:

lr_1, lr_2 and lr_small		for lawrencium
no_1, no_2 and no_small		for nobelium
md_1, md_2 and md_small		for mendelevium

Normally, only the first queue will accept job submissions. This is meant to prevent overcomitting of resources.

The subgauss script understands that saying "1" is requesting the lr_1 queue, for instance. The 1 and 2 queues are unlimited memory, unlimited disk, etc queues. The small queue is meant for running test jobs of short duration (less than an hour). You can see the queues by typing:

      qstat   -as

at the command prompt. It will tell you what jobs are queued on what queue at the current time.

The subgauss script will write the .log file in the same directory as the .com file. The .chk file will be automatically created in the same directory as well.

The Gaussian scratch files are automatically created elsewhere and deleted after the job completes (or fails). If there isn't enough space on /scr, or if you need to access more than 2 GB, please let Computer Support know; you will need to split the scratch files into a few files, since the RS/6000 Unix operating system (AIX) does not allow file sizes above 2 GB.

Gaussian normally defaults to 32 MB of memory per job -- this can be controlled by the %mem command in words; there are 8 bytes in a word (%mem=4000000 would be 32 MB of memory). Each of the machines has 128 MB of memory. We recommend that, at the beginning, you use the default memory and expand as you need only for particular cases.