- Intel Celeron G1610 (Ivy Bridge)
- Intel Desktop Board DH67BL
- 8 GB DDR3 RAM
- Hitachi Travelstar 7K1000 1 TB 2,5" (OS)
- Toshiba L200 2 TB 2,5"
- Hitachi Deskstar 5K4000 4 TB 3,5" (2x)
- Seasonic G-360 80+ Gold
- Scythe Big Shuriken
- Corsair Carbide 330R case
Originally running Arch, then Debian Wheezy, now Debian Buster pre-release in search of an ever lower idle power consumption.
See the powersaving tweaks below for more information. The server is scheduled to shut down in the morning and powers up again in the evening. The storage disks are ZFS pools with mirror disks in eS-ATA cases (offline, but not off-site), ZFS code is from the ZFS On Linux project. Disk health is monitored through smartmontools, which performs regular checks, and e-mails me when something is wrong. The data disks get a scrub every two weeks; I get an e-mail if errors are detected. The server up can be brought online through Wake on LAN (yay for smartphones and tablets!).
- NFS v3 and v4 server. OpenELEC does not support NFS v4 yet (nor does libnfs, where Kodi gets its built-in NFS support).
- SSH for remote administration.
- MariaDB server for the Kodi database.
- Lighttpd web server with a status page reporting disk usage, daemon status etc. through a shell script.
- rtorrent 'server' with rutorrent web front-end.
- SickGear and Autosub instances.
The original system idled below 20W (19,4-19,7W to be precise - measured with a Voltcraft Energy Logger 4000). That was with one 2,5" and two 3,5" hard disks. I have applied the following tweaks to bring down power consumption:
- Disabled all USB ports except for 2. Does not seem to impact idle power consumption a lot.
- Disabled the on-die GPU and onboard audio.
- Powersaving CPU governor enabled by default.
- Hdparm spindown. The 3,5" data disks go into standby after 10 minutes. The OS 2,5" disk does not seem to like hdparm's settings, going into standby every few minutes to be spinned up again after a few seconds. Spindown on that disk has been disabled.
- S-ATA ALPM enabled.
- Forced PCI Express ASPM. Seems to have little to no impact on idle power consumption; both on 3.2 and 3.12. The Intel board has a BIOS option for it (disabled for now) so it should be properly supported by the board, but the ASPM feature seems to be/have been a mess on Linux.
- Forcing ASPM does not seem to have any influence on the PCI-E buses' power settings. I enabled this manually. Saves less than 1W.