Hacker News Reader

Comment by sbierwagen | Hacker News Reader

  Seriously — laptop disks in a server?

  It might sound strange, but laptop disks actually have a lot of 
  advantages for an always-on home server: they use less energy, run 
  very quiet, and take up little space. Although you pay roughly 
  double the price per gigabyte compared to 3.5" disks, you pay less 
  for electricity, and thus save money in the long run.

I seriously doubt the power savings over the lifespan of a hard drive (5-6 years) are greater than paying twice as much per gigabyte. After all, if they were, Backblaze themselves would be using 2.5" disks.

In any event, the disks in a home storage server are going to be spending most of their time spun down, and drawing very little power.

View on HN

You are correct. Innumeracy raises its ugly head again.

Powering a 3.5" disk drive costs about $4/year.

Remember the rule of thumb: Each watt that you run all year long costs you $1/year in electricity.

(Also, Backblaze drives probably spend most of their time spun down and thus use nearly no electricity.)

EOM

Some data:

  Model      WD20EARS[1] WD10TPVT
  Size          3.5"         2.5"
  Cap            2TB          1TB
  R/W             5W         2.5W
  Idle          3.4W        0.85W
  Standby       0.7W        0.25W
  Cost           $83         $118

So, figure the drives are mostly idle, but spinning, you would be $4/year for the 3.5" and $1/year for the 2.5", but you might need twice as many of the 2.5" drives to get the storage if you really need 2TB, but if you don't, then it will be 10 years to recover the savings on electric, but that is your own fault for not figuring out how to spin down your drives, and not understanding opportunity cost, you can invest that $30 you saved and make $1.5/year forever, even after your drives detect the end of their warranty and fail in 3 years.

[1] Don't buy this drive. Something went horribly wrong in WD QC land. 1/3 of these that I bought failed their S.M.A.R.T. conveyance tests right out of the box and more failed in the first few months. The reviews at NewEgg suggest this is not an isolated experience. But, they are the ones I'm saddled with and know their data. Mine live out their lives happily spun down but for a few minutes a day and are closely monitored and mirrored, so their admittedly high risk of failure is acceptable.

For what it's worth, your figures for the power draw of the WD20EARs seem to be too low -- Storage Review% measures R/W 6.5W-7.1W, idle 6.3W (odd) -- while the 2.5in drive figures are about right, but maybe a bit on the high side.

And your rule of thumb for energy prices doesn't work in Germany (and possibly other countries), where 1W of electricity for one year will go for about USD 2.50, if not more.

% http://www.storagereview.com/western_digital_caviar_green_2t...

The numbers I used are from WD's spec sheet. The Storage Review discrepancy is too large to easily explain. Either something is wrong with that test, or WD has posted fictitious numbers.

The $1/year/watt rule is for the United States with a $0.12/kwHr cost. Even in the US the cost varies wildly, but just getting people within one order a magnitude on their electricity reasoning is worthwhile.

I had no idea German electric rates were so high, but it does help explain why large solar power projects are feasible there despite the less than exciting insolation.

I agree that the discrepancy is unusually high. The idle power consumption for other drives reviewed by them is lower and more in line with expectations.

Everything is more expensive in Germany, energy in particular. A rather high VAT (19%) in addition to taxes specific to energy, including gasoline and eletricity, cause this. Part of this is due to a conscious political decision to encourage energy conservation and investment into renewable energy; in fact, of the 0.2 EUR/kWh, about 0.03 EUR are a flat tax/allocation towards renewable energy. This has spurred photovoltaics, but particularly wind energy, which in 2010 accounted for 6.2% of the total electricity.

UK residents pay at least $2.80 per Watt per year according to my quick Google calculations.

On a different note, my last litre of petrol cost $2.12 (£1.33) - that's $8/gallon.

Given that electricity costs can vary by a factor of 5 or more from one place to another, that rule of thumb doesn't seem very, er, thumby.

I'm sure if one sampled the mass of human thumbs they'd find that it varies by more than 5.

$1/watt/year is based on a typical $0.12/kwHr US electric rate, but even that varies by a factor of three depending on location and usage category.

Yeah. So I looked it up for everybody. Wikipedia: http://en.wikipedia.org/wiki/Electricity_pricing#Price_compa...

Says 11.2ct/kWh in the US per 2011 (close enough to the $0.12/kWh cited elsewhere in this thread).

                $/kWh     $/W year
  --------------------------------              
  USA           11.20     0.98
  Germany       30.66     2.69
  Belgium       11.43     1.00
  Netherlands   34.70     3.04
  UK            18.59     1.63

You can calculate prices for other countries by looking it up on the table on Wikipedia and using Google Calculator to convert to dollar per Watt year:

https://www.google.com/search?q=($0.1859+per+kilowatt+hour)+...

Additionally, I propose that we redefine the name for the unit of "dollar per Watt year" as one Belgium.

---

Aside: is Wolfram Alpha ever useful for anything? Every once in a while I think of a question that would be perfect for WA to answer, and every time it disappoints me and I have to get the data from Wikipedia or Google, when WA could have also done the cost-per-year conversion for me straight away. This time it failed because as soon as you mention "energy price" WA assumes "US energy price" and tells me it has "no data" for the US energy price in Europe (which is technically correct, of course).

Addition: I forgot to do the price comparison for when it pays back the more energy-efficient harddisk versus the other one. Can someone do that?

How many Belgium do you need for the energy-efficient disk to pay itself back within warranty?

I'd do it even if it's about the same or just slightly more expensive, just for the sake of saving energy.

> Additionally, I propose that we redefine the name for the unit of "dollar per Watt year" as one Belgium.

I approve.

Note that lumping the entire US into a single number misses the fact that prices vary enormously within the US as well.

There's a huge table of prices in the US here: http://www.eia.gov/cneaf/electricity/esr/table6.html

The variation is amazing. While most of the prices are around the average you mention, some are quite low, with a decent number of people paying under 4 cents/kWh. On the other side of things, on place in Alaska pays over 100 cents/kWh, and a lot of others are well above the highest figure on your table. For somebody who buys power from the Middle Kuskokwim Elec Coop Inc, 1W costs $9/year.

Just for fun, here are some power consumption figures (based on data from storagereview.com%):

  Idle power draw (3.5in drives): around 5W
  Load power draw (3.5in drives): around 6.5W
  Spinup power draw (3.5in drives): around 15W

  Idle power draw (Samsung 2.5in): 0.58W
  Load power draw (Samsung 2.5in): 2.53W
  Spinup power draw (Samsung 2.5in): 3.89W

Of course spun-down (as opposed to idling but spinning) HDDs draw far less power. I found a 2008 report# which measured spun down power, and it ranges widely but 10 to 20% of idle seems to be about right. (Not sure if there's a 2.5in drive among those they tested; I only skimmed the report.)

If you disregard the spun-down thing for a moment, you get about 4W savings all the time, ie. 20W for the full array. That's about 175 kWh over a year. Assuming 0.20 EUR/kWh (which seems fairly typical for Germany), that's 35 EUR you save, per year. I guess that number can serve as an upper bound for the savings you can expect, chances are it'll be much lower: not a full array, 50% or more time spent spun-down, your energy prices might be lower.

% http://www.storagereview.com/samsung_spinpoint_m8_review http://www.storagereview.com/western_digital_caviar_green_2t...

# http://www.cl.cam.ac.uk/~acr31/pubs/hylick-harddrive2.pdf

>you get about 4W savings all the time, ie. 20W for the full array.

It would either be ~2W or ~7W. Relevant to bear in mind the TB 2.5inch drives have more platters than standard height 2.5" so will use more power.

5x2.5" Drives = 2.9W

1x4TB Drives = 5W 2x3TB Drives = 10W

You think 'i need X TB', not 'i need X drives'.

The other advantages he mentioned (reduced heat, vibration, and as a consequence noise) would outweigh the price difference for a "living room" storage server.

I certainly wouldn't want multiple 3.5" HDDs in my living room...

The disks are quiet, but the cooling fans they often require are not (at least that's what I've learnt from having a Drobo with four 3.5" disks in my living room).

thats why i use an ultra-silent 140mm fan ... running it at 5V (instead of 12V) makes it practically unhearable.

Why do hard-drives, in a small home server, need cooling fans?

They probably don't.

http://labs.google.com/papers/disk_failures.pdf

yes, that is true, the disks will run fine without. airflow is mostly for any heat generated by the sheeva and port-multiplier boards .... just to be on the safe side.

well, it partly depends on the local electricity rate you are paying .... which is probably a little higher here in europe than in the states, i guess. haven't really calculated the amortization time yet, though.

10 watts per year is 87.6 kW h... how much are you paying for your electricity?

88kwh cost about 22$ here in Vienna, AT.

So for a 200$ drive versus a 100$ drive using 4x as much power with both having an expected life time of 4 years, the 100$ drive is still cheaper.

Energy costs PER GB can actually be lower/much closer in a 3.5 inch HDD by going with the largest capacity drives.

Perhaps power saving is the point, not money saving?

Reply on news.ycombinator.com