[Rack] Power / Stallion Replacement
ben at nerp.net
Sat Sep 17 18:09:54 UTC 2011
Ok, I'll bring that SSD device in and see if we can get the ITX board to
boot off it.
"UNIX is user-friendly, it's just picky about its friends."
On Sat, 17 Sep 2011, Jonathan Lassoff wrote:
> On Fri, Sep 16, 2011 at 8:36 PM, Ben Kochie <ben at nerp.net> wrote:
> I did some power testing in the rack today.
> Stallion (2x Xeon 3.4GHz, 3GB ram, ~271 GiB SCSI disk raid) is using about
> 300W. I figure the pony machine is about 150-200 (I need to power it down
> to test). The rest of the stuff fills out the rest to total the rack
> power at around 650 Watts.
> This is contributing to about 450kWh/month in power (~$100) This means
> that stallion is using about $50/month in power. The tahoe node machine
> has something like 6G of ram and 6T of disk. We could easily run stallion
> and pony on it as VMs and cut the rack power in half. Or we could just
> replace stallion with that machine and still save $25/month in power.
> I think that one of the design goals of setting up stallion was to have a host to support things
> that really ought to be up all the time (like the touchscreens, or front gate relay).
> However, I'm not sure why it ended up in such a beefy machine just to host NFS mounts and TFTP.
> As previously suggested, the small mini-ITX board that has a bunch of serial ports could easily
> host NFS and TFTP with some more storage.
> As for the LAFS box, I wonder if the security of the host would be impacted by either running in
> or hosting a VM on it. I suspect that if we're just hosting a storage node and not a gateway,
> either would be fine.
> Is the LAFS box already imaged up and ready to go? If so, it might be easier to add a VM to the
> existing system.
More information about the Rack