On Fri, Jun 18, 2010 at 11:04:46AM +0100, Ed W wrote: > There is a guy who does Mac Mini hosting which seemed like an almost You probably mean http://www.macminicolo.net/ > interesting idea. They list at a price more expensive than a bunch of > supermicro Atom machines, but given that they use the 25W core2 part and > have been really built for low power it seems like you might gain quite > an advantage in power/cooling costs? They also seem to have a high Not really. The Mac minis have no IPMI, do not integrate well into racks so you'd need shelves, which means you'd have to build your own data center. I also doubt they beat Atom in OPS/J. Back-to-back short rackmounts ventilate well by convection (smoke funnel effect). In terms of performance/price they're not really that good (and have gotten worse with the latest Mac mini). They do not have 24/7/365 specced hardware (notebook hard drives). Etc. > density and of course if you are CPU bound rather than RAM bound then > the Core2 is likely vastly faster than an Atom? If you're CPU bound, you just buy more cheap boxes, and charge power users more. > I can only assume that these boxes survive reasonably well in a hosting > environment because someone else is already doing it... I would like to see their failure rate. I presume the main reason to run Mac mini is to cater to OS X users on the cheap. The OS X server Mac mini is even less attractive in terms of price/performance. > Anyway, so many ways to skin a cat... There are many feline species. My target are people who value a thin FLOSS wrapper about raw physical resources (address space, bandwidth, computation, storage). Proprietary components would be only in the way here. > Ed W -- Eugen* Leitl <a href="leitlhttp://leitl.org">leitl</a> http://leitl.org ______________________________________________________________ ICBM: 48.07100, 11.36820 http://www.ativel.com http://postbiota.org 8B29F6BE: 099D 78BA 2FD3 B014 B08A 7779 75B0 2443 8B29 F6BE