Summary on Powering Down Machines

Robert Lee rlee at digigw.lab.nanko.digital.co.jp
Tue May 29 19:03:50 AEST 1990


Thanks to everyone who responded to my original posting on powering
machines off and on everyday.  The responses were the type I was looking
for and helps me out alot.  Here is a short summary:

- with 1 exception, all the respondents left their SS1s and servers
  running 24hrs/day.
- the one person who turned off the machines each day, has had NO
  problems.  
- effectively, almost everyone said that it is better to keep the
  equipment at a constant temperature.  Here is an excerpt from
  lehners at uniol:

> The technical reason are also clear to me: the electrical, mechanical
> und temperature stress is really bad when switching on and off all the
> days. The peak currents at power on are real high (aprox. 5 times the
> current in normal mode). Also the temperature stress and the resulting
> mechanical stress is bad for the machines. Hard disk drives are really
> in stress when accellerating to the operating speed.

    - in addition, bob at kahala.soest.hawaii.edu said:

> ... Machines that are turned off and
> on every day definitely suffer hardware problems more often.
> This is particularly noticeably where humidity is high
> (leaving machines off at night can cause small amounts
> of moisture to condense, which leads to corrosion of
> contacts).

    - pallen at atc.boeing.com made a good point which is worth mentioning:

> Suns are designed for continuous operation.  We have all our machines on
> Sun maintenance and run them 24 hours a day.  The only moving part that
> seems to wear out is cooling fans.  They last a year or two and then either
> stop turning or start making noise.  Fans should be checked periodically.
> The other things that go wrong, like monitors and disks, would probably
> have more problems if the machines were power-cycled daily.  But then I
> don't have any hard data to back up that opinion.

- and finally thanks to p.mellor at uk.ac.city for his informative readnews
  article.  I will go looking in the archive for old Sunspots issues.

Robert
Internet: rlee at digital.co.jp



More information about the Comp.sys.sun mailing list