Power Sources and Their Coming Importance To Your Business

Do you have your own website? If you do it’s hosted on a server. Do you know where that server is located? Do you know the type of carbon footprint that server has where it is hosted? Do you care? If you do pay attention and you’ll learn something.

The first step if you do not know the actual data center where you server is located is to ping your domain name to get your ip address

So for example if I ping rackforce.com I get

Z:\>ping rackforce.com

Pinging rackforce.com [209.97.208.205] with 32 bytes of data:
Reply from 209.97.208.205: bytes=32 time=1ms TTL=253
Reply from 209.97.208.205: bytes=32 time=1ms TTL=253
Reply from 209.97.208.205: bytes=32 time<1ms TTL=253
Reply from 209.97.208.205: bytes=32 time=1ms TTL=253

Ping statistics for 209.97.208.205:
Packets: Sent = 4, Received = 4, Lost = 0 (0% loss),
Approximate round trip times in milli-seconds:
Minimum = 0ms, Maximum = 1ms, Average = 0ms

If you already know your IP address you can skip the above and go straight to the next step. In your web browser go to ARIN and put in your IP address in the Search WHOIS bar upper right hand corner of the page. If it returns and tells you that the IP is not in ARIN but is in APNIC, AFRINIC,RIPE, or LACNIC go there and repeat the process but what follows is for ARIN only – that is North America)

Putting 209.97.208.205 into the WHOIS form returns

OrgName: RackForce Hosting Inc.
OrgID: RACKF
Address: Suite 210
Address: 1628 Dickson Avenue
City: Kelowna
StateProv: BC
PostalCode: V1Y-9X1
Country: CA

NetRange: 209.97.192.0 – 209.97.223.255
CIDR: 209.97.192.0/19
NetName: RACKFORCE-2
NetHandle: NET-209-97-192-0-1
Parent: NET-209-0-0-0-0
NetType: Direct Allocation
NameServer: NS1.RACKFORCE.COM
NameServer: NS2.RACKFORCE.COM
Comment: Rwhois pending.
RegDate: 2004-05-07
Updated: 2005-01-27

and you can see by that the IP Address 209.97.208.205 is hosted at RackForce. Now, on to the meat of this post and why we dallied so to begin with.

If we look at energy, first we will see how the source of that energy is important when considering the carbon footprint of a data center.

Data source: Parliamentary Office of Science and Technology – UK &* European Data – 2004-2006

While wind/solar energy are obviously the best (Nuclear is good, as far as carbon is concerned, but there are other problems with it as a source of energy) but they are not reliable as there is a problem with energy storage for when the wind is not blowing, or the sun is not shining. Efficient data centers require a very large amount of power (~40MW) and that power has to be continuous, without fluctuation. That leaves hydroelectric power as the best source of power for a data center. Right away, you can see that if a data center uses hydro instead of coal it has approximately a 99% reduction in its carbon footprint.

Of course getting 100% pure hydro-power is difficult. Unless it connects directly to a dam then a data center is on one of the many electrical grids out there in which case it is getting a mixed power feed. That needs to be considered when calculating your footprint. When you do, you get a picture (for Canada and the US) like this

Take for example British Columbia in Canada (green arrow on the left). Almost exclusively hydroelectric but does use some natural gas for peak time generation. Thus instead of being around 5 gCO2eq/Kwh (grams CO2 equivalent per kilowatt hour) it is approximately 17 gCO2eq/Kwh, then look at West Virginia which is 100% coal powered (the US power grid is 50 to 52% coal generated) and it is 1055 gCO2eq/Kwh. Where would you rather get the juice from that powers your ICT equipment? Green locations, 20 grams CO2 or less range, 200 or more is extreme – 10X worse.—starts just a little ways into chart. The source of power becomes even more important when we start increasing the efficiency of data centers. In economics, there are principles (Khazzoom-Brookes Postulate) that state that improvements in energy efficiency work to increase, not decrease, energy consumption by reducing the cost of the amount of work achieved for energy spent. Thus if you make your data center super efficient it will lead to greater power consumption2 and instead of reducing your footprint you will increase it and the dirtier your source of power is the worse your footprint will be. However, this must not be seen as an argument against efficiency, there are other tangible benefits that flow from efficiency, but rather as an argument for selecting the greenest power source fueling that efficiency ¹

Let’s look at an example of two data centers, one in West Virginia and the other in British Columbia. Based on the data from Stats Canada, Environment Canada, & US Department of Energy that I researched I was able to build a spreadsheet showing the likely carbon cost for operating a server in each province and state (click on image below for a readable version)(terms: gCO2eq/Kwh= grams of CO2 equivalent per Kilowatt hour, mTCO2eq/Mwh = metric Tonnes CO2 equivalent per Megawatt hour, PUE = Power Usage Effectiveness a way of measuring how efficiently a data center uses the incoming power, that is what is the ratio of power used by the data center to the amount of power required to operate the ICT [Infornation Communications Technology] equipment (servers, switches, routers) – 1:1 would be perfect but basically impossible)

Now let’s see what that could mean to your business.

Say each data center is 120,000 sq. ft. raised floor (not at all unusual) Now allow a standard 32 sq. ft. per cabinet. That would give you a maximum number of 3125 racks (120,000/32) and each rack can hold a maximum of 42u worth of gear (a standard rack) but some of that will be the power distribution units of the data center and likely some of their networking gear too, so in general you will get around 36u of usable space. Assume you put 36 1u 200W servers in those slots. That gives you 112,500 servers in those 3125 racks. In BC, each of those servers would cost you an additional $1.06 per year. In West Virginia that would be $65.72 extra per server (the actual results would be higher though as a 120K sq. ft. data center would use up at least 20% of that space on aisles and various components needed to run a data center.) which, translates to $2365.92 instead of $38.16per rack per year extra. How will you justify that extra $2327.76 per rack per year to your shareholders?

The calculations above though were theoretical. They were based on a data center with perfect utilization of energy. That is for every watt of power required to run the ITC equipment in that data center they only used 1 watt of incoming power. Sadly, that is not the case and the average data center today has a Power Usage Effectiveness (PUE) rating of 2.5 (and many are much worse – that is an average). That means they need to purchase 2.5 watts of power for every watt they sell to their customers. Now go back to the last paragraph and multiply those final numbers by 2.5. Your extra cost is now $4654.52 per rack.

If your company is a public company then, as carbon taxes and/or carbon Cap &Trade becomes legislated then you will have a fiduciary responsibility to your shareholders to seek out the option that has you paying the least amount of taxes in order to maximize your returns. If you are a private company then you still need to consider the source of what powers your servers lest your competition beats you to it and gets a substantial edge in costs over you.