[LINK] Leave on, or turn off?

Ivan Trundle ivan at itrundle.com
Thu Nov 30 16:29:29 AEDT 2006


I still think that there are too many imponderables, and too many  
generalisations. To illustrate this, try substituting  'automobile'  
for 'computer' or 'server'.

We expect vehicles to last a period of time (generally measured in  
years, sometimes kilometres travelled), but expect that (anecdotally  
- I'm not pushing any barrows here) Toyotas are built better than  
Fiats, even though the underlying components are much the same. Usage  
patterns vary wildly, too - and yet we still make these assessments.

I suspect that computers are much the same. A server (one would hope)  
is *designed* from the outset to work continuously, over a period of  
time. Laptops are built to survive the rigours of movement,  
environmental thermal variation, being put to sleep and so on.  
Desktop machines are designed to work in a different, more stable  
environment, but not perhaps with the same demands as that of servers.

More specifically, hard disks are designed with bearings that are  
expected to spin for a period of time before failure (marketing  
people aside, there are engineers that specify these details before  
the marketing dept get to run over the data), platters are designed  
to spin for a period of time, arms that are expected to transport the  
read and write heads for a period of time, and logic boards that are  
expected to cope with electrical input for a period of time.

On the hard disk alone (just one of many components under the  
spotlight here), you might want to read this:

http://www.bitmicro.com/press_resources_war_of_the_disks2.php

Horses for courses - the different speeds and capacities of hard  
disks, etc, all need to be factored in.

And on an anecdotal level - I can attest to having a fanless desktop  
machine that has run continuously since 2000, acting as my home  
server. I have other servers both at home and at work that have run  
for longer, with no faults - and have withstood power brown-outs,  
surges, lightning spikes, and the regular temperature fluctuations  
that a non-airconditioned space affords.

I suspect that Stewart is alluding to the overall cost in having a  
machine either on or off at times, but factors such as power spikes  
and other malevolent events depend on how well the box is constructed.

Going back to the analogy of cars:

You wouldn't expect a Hyundai to take you across Africa, but then a  
Toyota Landcruiser might just be over-engineered for the daily trip  
to and from school in North Sydney. But, ultimately, they are all  
designed for a particular task, and with a particular life cycle.

iT


On 30/11/2006, at 3:27 PM, Stewart Fist wrote:

> Thanks to all for the rash of anecdotes -- they have helped me  
> clarify the
> question, although I'm still dubious about many of the answers.
>
> Do people accept that a manufacturer's MTBF is something devised by  
> rigorous
> testing rather than a figure invented by the only mathematically- 
> literate
> spin-doctor in the PR department.
>
> Turning off the monitor (especially CRT) is obviously a good  
> thing.  I'm not
> sure how much of the monitor electronics are active during sleep mode.
>
> Also for home computers, especially when users are at work, the  
> risk of
> leaving the machines linked to powerlines during thunderstorms must  
> be a
> major consideration.  In pure monetary terms, this probably  
> outweighs any
> concern about hard-disk bearing wear.
>
> A couple of other sub-questions:
>
> 1. If a PC goes into sleep mode, turns down its hard-disk, and cuts  
> power to
> its monitor, wouldn't it suffer pretty-much the same thermal shock  
> when it
> powers up ?
>
> Doesn't that kill the thermal-shock argument?  Although, as Ivan  
> points out,
> most modern machinery is designed to accommodate this.
>
> 2. How do servers stand up against desktop PCs in terms of long-term
> reliability?  Are differences just due to more attention to cooling or
> better components, or is there some other factor involved ?
>
> 3 Do hard-disks wear out faster when they are constantly  
> accessed ?  Or is
> the life-span set purely by running time, irrespective of actual  
> usage.
>
>
> -- 
> Stewart Fist, writer, journalist, film-maker
> 70 Middle Harbour Road, LINDFIELD, 2070, NSW, Australia
> Ph +61 (2) 9416 7458
>
> _______________________________________________
> Link mailing list
> Link at mailman.anu.edu.au
> http://mailman.anu.edu.au/mailman/listinfo/link
>
> !DSPAM:1009,456e5e5f420303736230347!
>
>

--
Ivan Trundle
http://itrundle.com ivan at itrundle.com
ph: +61 (0)418 244 259 fx: +61 (0)2 6286 8742
skype: callto://ivanovitchk





More information about the Link mailing list