This rack also contains (not yet on managed PDUs):
* One more of the same model 40/100G switch (so likely another ~40W)
* 3x old Cisco 24-port switches that I expect are major power hogs and plan to...
As of my last power bill, my effective rate after taxes and fees was $0.217/kWh.
So ignoring contributions from solar as well as non-unity PUE, the annual power OPEX for the identified equipment is:
* Ceph cluster: $545
* Network: $315
* Backup server: $400
* Rest of the lab: $6343
The backup server seems like an easy target. I can't get rid of it, the deal I have with my friend is that I host her offsite backups and she hosts mine and no money changes hands. But it's full of old enterprise SAS drives that are totally not necessary for something whose performance is dominated by internet latency/bandwidth and I am more than willing to buy some hardware upgrades to make it more efficient over the long term.
The current network stuff I don't see changing much, I need high bandwidth and am willing to pay for it. The old Cisco switches are likely much more power hungry than the currently identified loads and I already have a migration plan to get off them (eventually).
The storage cluster is probably also not changing.
So I guess the next step is to buy another managed PDU and collect more data, plus work on making the backup box more efficient.
So ignoring contributions from solar as well as non-unity PUE, the annual power OPEX for the identified equipment is:
* Ceph cluster: $545
* Network: $315
* Backup server: $400
* Rest of the lab: $6343
The backup server seems like an easy target. I can't get rid of it, the deal I have with my friend is that I host her offsite backups and she hosts mine and no money changes hands. But it's full of old enterprise SAS drives that are totally not necessary for something whose performance is dominated by internet latency/bandwidth and I am more than willing to buy some hardware upgrades to make it more efficient over the long term.
The current network stuff I don't see changing much, I need high bandwidth and am willing to pay for it. The old Cisco switches are likely much more power hungry than the currently identified loads and I already have a migration plan to get off them (eventually).
The storage cluster is probably also not changing.
So I guess the next step is to buy another managed PDU and collect more data, plus work on making the backup box more efficient.
1