Skip to main content

You may be using a Web browser that does not support standards for accessibility and user interaction. Find out why you should upgrade your browser for a better experience of this and other standards-based sites...

Dartmouth Home  Search  Index

Dartmouth Home | Search | Index

Dartmouth home page
Computing at Dartmouth
Computing > Support >  Library >  Research > Start >

Research Cluster Etiquette (Proper Use of a Shared Resource)

UNIX/Linux is a multiuser operating system. It is important that users of the Research Cluster understand how to work properly with a shared computing resource. When used properly, the Research Computing machines can meet the needs of many simultaneous users. But users who are careless can cause problems for others and for the computer.

To use the Research machines effectively, users need to know the answers to the following questions: "How busy is the computer?","What jobs am I running on the computer?", and "What impact are my jobs having on the computer?". Each of these questions is discussed below; links to the commands that answer the questions are provided.

How Busy Is the Computer? (Commands = uptime, xload)

Whether it is a line at the bank or a number of cars waiting to park in a lot, it is important to understand how busy a shared resource is before you make a decision to use that resource. One measure of how busy a UNIX computer is is the "load" metric. The "load" provides a rough measure of how busy a computer is and gives the user some idea as to what kind of performance they can expect. The Research Computing machines have 4 CPUs (Ozark and Polaris), and 16 CPUs (Andes), so dividing the load by the number of CPUs will provide a rough idea of how busy the machine is. General guidelines are provided below:

Load

What It Means

4 or under (Polaris)

Computer is OK to use.

4 and over (Polaris)

Computer is busy, but should be OK to use.

16 or under (Andes)

Computer is OK to use

16 and over (Andes)

Computer is quite busy; use another machine

For "uptime" information, see How Busy Is My Computer?

What Jobs Am I Running on the Computer? (Commands = ps, top, kill)

One of the more common problems that occurs on the Research Cluster systems is when a single user runs multiple jobs, but is unaware they have many jobs running. The ps command will display the jobs you have running, and the top command can be helpful if your jobs are active. The kill command removes unwanted jobs. In general, users should only kill their applications (GCG, fasta, SAS, matlab, maple, etc.), as it is the actively running applications that cause the most problems.

For information on the ps command, see How Do I Know What Jobs I Have Running?.

For information on the top command, see How Do I See If My Job Is Running?.

For information on the kill command, see How Do I Kill a Job or Process?

What Impact Are My Jobs Having on the Computer? (Commands = top, ps, uptime, xload)

There are a couple of different ways to "see" what jobs are running on the computer and the impact of those jobs are having. The top command is probably the best all around tool to use when you want to see what is happening on a computer. Top shows the most active processes (the jobs that are actually running and using up CPU cycles); it also provides resource usage metrics (CPU and memory). Like any other program, top uses system resources, so it is best not to leave top running all the time. Top can be very useful when used in conjunction with xload. One reasonable way to work is to always keep an xload window up on your display; if the load starts to climb, use top to see what jobs are impacting the system.

For information on the top command, see How Do I See If My Job Is Running?.

05/28/08

Last Updated: 5/28/08