Re: [Beowulf] RE: Capitalization Rates - How often should you replace a cluster? (resent - 1st sending wasn't posted ).

2009-01-20 Thread Mark Hahn
Think about it: if you pay 1/3 the cost of a new node per year in rent and utilities, that seems amazingly high to me. I don't have much sense for rent, but even if we ran our clusters on "retail" power rates it would only amount to 2.3% of purchase cost. you could pretty easily argue that ou

Re: [Beowulf] RE: Capitalization Rates - How often should you replace a cluster? (resent - 1st sending wasn't posted ).

2009-01-20 Thread Orion Poplawski
On Tue, January 20, 2009 12:42 am, Jon Aquilina wrote: > so basically all those old machines end up on ebay > Yeah, what's the capitalization vs. operating cost argument when the capital cost is about $200-$250? :-) I've come to really like eBay recently -- Orion Poplawski Technical Manag

Re: [Beowulf] RE: Capitalization Rates - How often should you replace a cluster? (resent - 1st sending wasn't posted ).

2009-01-20 Thread richard . walsh
- Original Message - From: "Greg Lindahl" >Hey! In year 4 it's about the same to keep the old cluster, or throw >it out and buy a new one. (If I want to spend the same $$, I can buy a >cluster 1/4 the size, but same performance, as the original one.) > >And in year 5, it's a big

Re: [Beowulf] Nehalem and Shanghai code performance for our rzf example

2009-01-20 Thread richard . walsh
- Original Message - From: "Bill Broadley" b...@cse.ucdavis.edu >If gallium arsenide or some other material gave us 10x the clock rate per >watt, but 1/2 the transistors would it really matter?  Seemed like even intel >is begrudgingly admitting it's the memory bus, and finally the

Re: [Beowulf] RE: Capitalization Rates - How often should you replace a cluster? (resent - 1st sending wasn't posted ).

2009-01-20 Thread Geoff Jacobs
Mike Davis wrote: > Lux, James P wrote: >> >> On 1/19/09 11:14 PM, "Jon Aquilina" wrote: >> >> >>> in all honesty is there a need really to replace the cluster. why not >>> just add >>> onto it? >>> >>> >> >> 1) If all the processors work at the same speed, it's "easier" to >> distribute >>

Re: [Beowulf] RE: Capitalization Rates - How often should you replace a cluster? (resent - 1st sending wasn't posted ).

2009-01-20 Thread Mike Davis
Lux, James P wrote: On 1/19/09 11:14 PM, "Jon Aquilina" wrote: in all honesty is there a need really to replace the cluster. why not just add onto it? 1) If all the processors work at the same speed, it's "easier" to distribute work among them, especially if there are synchronizati

Re: [Beowulf] RE: Capitalization Rates - How often should you replace a cluster? (resent - 1st sending wasn't posted ).

2009-01-20 Thread Lux, James P
On 1/19/09 11:14 PM, "Jon Aquilina" wrote: > in all honesty is there a need really to replace the cluster. why not just add > onto it? > 1) If all the processors work at the same speed, it's "easier" to distribute work among them, especially if there are synchronization steps periodically, wh

Re: [Beowulf] Module file for Intel 11

2009-01-20 Thread John Hearns
2009/1/20 Kilian CAVALOTTI > Hi John, > > > I would try to determine the machine type in the module file itself. Not > sure > if a modulefile could get the result of an external $(uname -m | sed > 's/x86_64/intel64/;s/i[3-6]86/ia32/'), but the modulefile(4) manpage > mentions > a uname directive.

[Beowulf] AMD quad-core recommendations

2009-01-20 Thread Eugen Leitl
This is the time of the year where I annoy the list with pointless questions again. I used to use an X2100 M2 Sun box as a staple, but for some unfathomable reason Sun decided to jack up the dual-core barebone price by 300+ EUR. Real smart move, that. I've used HP ProLiant DL165G5, but FSC VF

Re: [Beowulf] Module file for Intel 11

2009-01-20 Thread Kilian CAVALOTTI
Hi John, On Tuesday 20 January 2009 13:46:14 John Hearns wrote: > The recommended way to set it up is to source ifortvars.sh > It rather goes against the grain to have a module which works out the > architecture then sources this file If you have an homogeneous cluster, architecture-wise, you ca

Re: [Beowulf] Module file for Intel 11

2009-01-20 Thread Joe Landman
John Hearns wrote: Hopefully not too far off-topic. Does anyone have an outline modules (modules.sourceforge.net ) file for the Intel 11 compilers? The recommended way to set it up is to source ifortvars.sh It rather goes against the grain to have a module wh

[Beowulf] Module file for Intel 11

2009-01-20 Thread John Hearns
Hopefully not too far off-topic. Does anyone have an outline modules (modules.sourceforge.net) file for the Intel 11 compilers? The recommended way to set it up is to source ifortvars.sh It rather goes against the grain to have a module which works out the architecture then sources this file - a

Re: [Beowulf] RE: Capitalization Rates - How often should you replace a cluster? (resent - 1st sending wasn't posted ).

2009-01-20 Thread Robert G. Brown
On Tue, 20 Jan 2009, Jon Aquilina wrote: in all honesty is there a need really to replace the cluster. why not just add onto it? It depends on your resources. You have to provide electricity to the cluster nodes, cooling for them, space for them, and they cost money to maintain and administer