Monitoring My Energy Usage (Part 4)
The first three parts of this series talk about the hardware and software I have set up in order to capture my energy usage. Of course, it's a good fun (and somewhat geeky) project but ultimately the aim is to help me save energy. So now it's been up and running a little while, it has yielded results which have taken a slightly unexpected turn.
In part 1, I discussed the Current Cost unit, and briefly mentioned that it is plugged into my home server where all the software magic happens. What I didn't mention, or appreciate at the time, is that the server itself is quite an energy hog. Of course, the system was up and running 24/7 before I embarked upon this project, so I haven't just shot myself in the foot in that sense. However now that I've had a while to sit back and look at the readings, I have realised just how much energy that server has been using.
On days when I'm away from home (at work, or out for the day), I have discovered that there has been a constant load of between around 200 and 300W.
Needless to say, when I'm not at home, there are not that many things drawing power:
- Network (router, wifi AP, couple of switches, Drobo)
- Fridge/freezer unit
- A couple of clocks and items on standby
With a bit of experimentation, I've discovered that the server accounts for somewhere between 100 and 150W of that load. The other ~50W is the network and things on standby, and the intermittent extra ~100W seems to be the fridge.
It seemed to me that this can be improved on.
Knowing You Have a Problem...
The server as it exists now is a pretty hefty beast. It's an Intel Core 2 Duo based machine, with 3GB RAM and 2x 500GB HDs (in RAID 1). Already you can probably see where the problem is. At the time this seemed like a great idea and to be fair, it has done everything I've asked of it flawlessly. However I decided that the time has come to downsize it a bit and strike a compromise between ability and energy use.
... And Finding a Solution
First thing first, the two hard disk situation needs reviewing. Yes, having the security of a RAID 1 setup is nice, but that's instantly doubling the amount of power going on keeping them spinning. There's nothing so important going on for which a regular backup routine is not good enough.
The second thing is that the system itself (despite the nice "efficient" Asus motherboard) uses too much power. Instead of a big system like this, cooled my no less than three separate fans, I decided to go down the path of small and efficient and built out a new mini-ITX system.
Thanks to the Mini-ITX.com Store, I obtained the parts for a brand new, small and efficient little server. It is built around a Jetway JNC96-525 fanless dual core Atom board, and has a 2GB stick of RAM (with space for another 2GB if/when I need it). The board also takes a 12V DC power input, so no need for me to buy a PSU as it's already built right in. I just provide an adequate AC/DC power brick and it is all taken care of. In place of the two big hard disks, I now have one 2.5" drive (320GB) with a spindle speed of just 5400rpm – again, this will help to save power and provides enough performance for what I need of it. The whole thing is packaged inside a tiny M350 enclosure.
Fanless != No Fan
As you no doubt noticed, the entire system has no fans at all. It quickly turned out though that the board runs somewhat warm when inside an enclosure without any fans – too warm for my liking. Under stress, the CPU temperature (as reported by the BIOS at least) was topping 70°C. Apparently the board does run warm, but that is just too warm for me, given the fact this is meant to be running all the time. Thankfully, Mini-ITX.com also stock some very quiet 40mm fans (which fit the fan slots in the case), so I added one of those and configured the motherboard to throttle the speed according to the temperature. I still have it running warm, but far less warm than it was before, and it's still almost as silent.
Watt's Going On?
After all this, you're expecting it to use less power right? Me too – that was the entire idea. And thankfully, it does. It now sits somewhere between 25-40W, which is a big improvement. All in all, I'd call it a success. The only downside is that it seems the less power is drawn, the less accurate the Current Cost seems to be. Having said that, it is well known that this type of device is not 100% accurate, especially when dealing with certain types of loads. This is a compromise I knew about when I first bought it and set this all up, and it's one I'm happy to live with.