[LINK] Green web servers with netbook components

[Marghanita da Cruz]

Tom Koltai wrote:
> 
>> -----Original Message-----
>> From: link-bounces at mailman1.anu.edu.au 
>> [mailto:link-bounces at mailman1.anu.edu.au] On Behalf Of Adrian Chadd
>> Sent: Tuesday, 19 May 2009 11:37 PM
>> To: Bernard Robertson-Dunn
>> Cc: link
>> Subject: Re: [LINK] Green web servers with netbook components
>>
>>
>> On Tue, May 19, 2009, Bernard Robertson-Dunn wrote:
>>
>>> My experience of servers in data centres is that reliability and
>>> availability are very important, hence redundant power 
>> supplies, network 
>>> connections, SAN connections and separate networks for 
>> management and 
>>> back-up.
>>>
>>> PCs and PC motherboards don't really fit into this infrastructure 
>>> model.
>> Only because the infrastructure model that almost all 
>> companies these days are believing is what came out of the 
>> microcomputer industry and NOT what has come out of the 
>> mainframe computer industry.
>>
>> Funny, really...
> 
> Actually, Adrian has a point.
> Nano CPU's can be bonded for additioanl load.
> 
> Mercedes make a Deisel that turns off cylinders not in use.
> 
> It makes sense to devise a server that is capable of doing the same
> thing.
> Interesting - Bernard claims multiple cpu's are required for redundancy
> and load, and Adrian claims that a server with multiple cards with CPU's
> being turned on by demand is a more efficient model.
> 
>>From my CCI (Harris Scarfe) mini computer days (1985-86), I vote for
> Adrians idea. Smaller CPU's load sharing as demand increases; and one
> owould think, the individual CPU's can be selected to replace failed
> cards so redundancy is built in.
> That was pretty much the ICL DRS 300 SCSI (1990) backplane approach - it
> worked - well. Small paralell independent processor units that could
> load share tasks.
> 
<snip>

My hobby horse, from my Mainframe (Cyber 76 days), is lack of memory not CPU 
cycles. In those days we had to do manual memory management. A few years later, 
even though you didn't have to do memory management in a Vax cluster, I 
demonstrated that what was needed was need to speed things up was more RAM - not 
another CPU. I still maintain that most PCs need more memory (particularly with 
bloated operating systems) - than the stock standard configuration. Maybe that's 
why my laptops last 7 years - my current model has had two upgrades in sw XP 
 >Knoppix 3.2>Knoppix5.1 and is probably due for another (think about that from 
a green perspective)!

Ofcourse, memory is expensive and CPU processing is relatively cheap.

With regard to redundant power supplies - is this need based on evidence or just 
academic speculation. Sometimes redundancy (power supplies, raid disk drives), 
always at greater cost, results in reduced performance/reliability - but I might 
be out of date on this.

Marghanita
PS...and because yet again, someone feels the need to use a car Analogy we might 
as well check out <http://www.lexus.com.au/hybrid/models>
-- 
Marghanita da Cruz
http://www.ramin.com.au
Phone: (+61)0414 869202