|
|||||||||||
|
Re: [OT] low energy server HW
From: Anders Breindahl <skrewz(at)skrewz.dk>
Date: Wed Feb 20 2008 - 14:33:12 EST
That's some solar panels. You're saying that you have 52kW available at best? -- why are you worrying about server energy consumption, then? :) > > thinking about something like via c7 1.2-1.5 Ghz with 1-2 GB RAM Something along the lines of that would probably give you good performance for the money. Either that or AM2-cpu's from AMD. It's incredible how 2005 was 115W idle and 2007 was 80W idle. I would be considering the pico-ITX boards, too. Simply put, smaller boards tend to have lower requirements. For example the following, which is rather sweet (wonder if one could make a terminal server from such a thing?): http://www.mini-itx.com/2007/12/10/via-launch-artigo-pico-itx-builder-kit There exists some pretty hefty Core 2 Duo boards in that form factor, too. They, like AM2-cpus from AMD generally deliver scalable performance vs. energy consumption. > I like to change the last 4 drives against the same type like the If energy consumption of hard drives is important, consider flash-based storage, which are an order of magnitude more economical. (But take a stand with regards to flash outwearing). Actually---how mature do you, the debian-isp folks, consider flash-based storage for server usage? Obviously, one would shred the flash cells for files like mailspools. (I suppose the inode tables also will suffer, especially when using maildir, where files are created all the time). On the other hand, many on this list use battery-backed RAM (either by tmpfs+UPS or those I-RAM things) for peak loads on the file system anyway. So that only leaves somewhat more regular file systems usage to the disk, and coupled with kernel caching, SW/HW raid and firmware-level wear levelling ... It's rather reasonable, no? > I think, for a small Enterprise like mine or your project, the VIA At least earlier (I'm referring to C3, their P3-lookalike) VIA's CPUs were almost cache-evacuated. Their only performance-mitigating effect was that of the padlock instructions (3-5 orders of magnitude increase in AES performance (!)), but for general-purpose computing, they suffered from the little amount of cache. IMHO. > Even my new Workstation is a LN10000EAG (1GHz with 1 GByte of memory) Homemade generally is unreliable (and I'm not the *worst* solderer there is!) And what's more: You're the responsible, when things go wrong; you can't delegate it. > Note: The 15" TFT was consuming arround 86 Watt by 230V AC and now That *is* significant. I can sometimes get infuriated with how producers go for cheap and simply waste bunches of energy in conversion. Your figure is however abnormally gross. Good luck living in a container! :) Regards, skrewz. -- To UNSUBSCRIBE, email to debian-isp-REQUEST@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.orgReceived on Wed Feb 20 14:43:16 2008 This archive was generated by hypermail 2.1.8 : Wed Mar 19 2008 - 06:53:13 EDT |
||||||||||
|
|||||||||||