White noise as a system monitoring aid? Real or myth?

Question

Ok, I don't know if this is/was a sys admin urban myth, but when I was a student I heard of technique to help sys admins monitor a server without actively and constantly checking on it.

What they would do is route something like CPU usage or network saturation to a white noise generator. They would play this in the background at low volume, and if some unusual activity occurred the noise would sound different (volume, tone?) They would then check on the server using the regular console tools.

My question is, has anyone tried this? Does such a program or technique exist?

Answer 1

Real

This was certainly possible, and used, on at least some makes of older mainframes (1970s, early 1980s). The console (a Teletype, if you have ever seen such a thing) had a small speaker and a volume control. When this was turned up the audio output was indicative of the work the CPU was doing. Much of the work was off-line batch work and with practice it was possible to recognise certain distinctive 'signatures' of some jobs in normal mode and the shrill whine of a job in a loop. As the console would become sluggish when the machine was very busy the main use of the speaker was to allow the operator to check whether a job was progressing (the sound would usually vary) or stuck. On the lighter side, it also allowed jobs to be written to create certain signatures, for example a, somewhat atonal but recognisable, "Jingle Bells".

Answer 2

During the LISA 2000 (Large Installations System Administration) conference, Michael Gilfix and Alva Couch presented a paper: Peep (The Network Auralizer): Monitoring Your Network With Sound (PDF). While this research used jungle sounds instead of white noise, the idea is the same.

Answer 3

Solaris has an option to snoop ( the equivalent of tcpdump ) which produced some audio when a packet was sent or received. This was used so you could hear the network.....

Answer 4

I played with this a while back to "listen" to my mysql servers pagecache utilization : ) http://xtat.rapidpacket.com/blog/posts/what_querycache_sounds_like/

Answer 5

Don't use this on your production system. Just use Nagios. Please.

Answer 6

This seems like a really original technique to me :)

However, while being original, it is also a bit unpractical. What if I am working with my headphones on? You can use SMS to notify people, responsible for particular hardware/software. You can xmpp, or, hell, even email notifications.

Still, very original.

Answer 7

I've tied perfmon/wmi to SAPI ("Danger Will robinson!!) and I've tied counter alerts to kick off media player but never to w white noise generator. This really only works if you have exception based reporting set up, but it's fun for a while.

Answer 8

I first heard this story in 1985 from a couple of sysadmins. It was and "old" story then as well. Slightly different details, but still the same story.

I guess if you are alive long enough, all things will be new again. Now where did I put those bell bottom pants.....

Answer 9

Well as someone who works in datacenters, I have found that for severe problems white noise level changes are something you can detect almost subconciously. So when something starts overheating you can hear it and find it pretty quickly.

The problem is that when something starts making less noise than it was, it can be hard to find (ie quick -- find the quiet!)

Answer 10

not white noise, and i dont have a linux box to test with now, but

cat /var/log/syslog > /dev/dsp

or some other log

or i was thinking about using beep and feed the cpu load into it to change the frequency