The Hot Aisle Logo
Fresh Thinking on IT Operations for 100,000 Industry Executives

My friend, the inventor and green innovator, Peter Hopton from VeryPC has a unique view on how to be environmentally responsible whilst deploying PCs into an office environment and he is presenting his ideas at The Carbon Footprint Energy Efficient IT Summit on the 4th and 5th September.

Peter says that there are three basic approaches:

  1. Give each user their own PC and network connection (ordinary PC’s and Green PC’s optimized for energy efficiency)
  2. Give each user a Thin Client (dumb terminal and use VDI, Citrix or Windows Terminal Server from the Data Center)
  3. Use a shared and virtualized local office based computer serving multiple users (VeryPC GreenHive)

Calculating the per user energy usage of the solutions is pretty straightforwards once we take all of the variables into account:

  • Green PC Power Consumption p.a. – 21.1 MWh (£15,100 over four years)
  • Ordinary P4 Desktops p.a. – 49.7 MWh (£35,800 over four years)
  • Thin Client Power Consumption p.a. – 36.3 MWh (£26,100 over four years)
  • GreenHive Power Consumption p.a. – 4.9 MWh (£3,500 over four years)

Not what you’d expect?

  • It demonstrates the impact of burden on a model, had burden been excluded the model would have preferred the thin clients.
  • In fact, in this model for the thin clients to be on par with the GreenPC, a PUE of 1.5 is needed in the data centre.

Here are the key measures:

  • The power consumption of the Personal Computers and Thin Clients themselves. – Watts
  • The Watt/VA relationship – Power Factor ratio
  • The power consumption of server infrastructure required to support the Thin Clients. – Watts
  • The overhead of cooling and power protection infrastructure in the data center – PUE ratio

The Green Grid use PUE to measure the burden of servers on infrastructure in a data center. PUE doesn’t really work for desktop computers as in most climates the heat given off by the infrastructure is useful and works to reduce the need for heating of the office space. Peter introduces a concept called Burden Factor (Bf) that incorporates the concept of useful heat. For DCs with no useful heat PUE = Bf. (You can get more information on Burden Factor in this PC Pro VeryPC article.)

What Does This Mean?

  • For a Data Center based server, power consumption is very important as the supporting equipment will multiply its effective consumption (by the PUE).
  • A room with a high power density, and lots of supporting equipment will have a high PUE/Bf
  • An office will have a low Burden factor.
  • Do power calculations before you move processing cycles to the data room. 

Peter claims his understanding of Burden can be used to Improve the energy performance of thin clients

  • Locate your terminal servers at low density in your office.
  • Collaborate desktops in smaller clusters, so the terminal servers are easily turned off when not used, and a failed unit can be tolerated for 30 mins-1hr.
  • Save energy though using uncompressed short range data (client to server).

Using his understanding of Bf, Peter developed GreenHive

  • Instead of a 50 user 8 core server, a smaller, more efficient, 7 user dual core ‘Hive’
  • No compression, clients connected by KVM/CAT5 media adaptors
  • Use software/users to switch Hives off when not used.
Here are a few other takeaways from Peter:

Comparing Server costs based of energy use?

A popular dual quad core, modern big brand server is virtualised out, it will consume 300W at 80% load. At a PUE of 3, and a cost of £0.18/KWh, this server will cost £5676 to run over the next 4 years, much greater than its capital cost.

It will cause 13.5 tons of CO2, more than a typical 4×4 would over the same period. You’d have to have 242 living trees to absorb this much CO2 over this timescale.

An efficient dual quad core server will consume 140W at 80% load. At a PUE of 3, and a cost of £0.18/KWh, this server will cost £2649 to run over the next 4 years, still greater than its capital cost. It will cause 6.3 tons of CO2.

The saving is almost £3000, yet both pieces of equipment have similar computing power.

Supporting Notes

Electricity Costs and CO2

  • For modelling costs over computer lifetime choose an average electricity cost for that lifetime.
  • Today’s price after CCL is 12p/KWh, but over the last 4 years has doubled. So projecting the trend forward an average for modelling should be 18p/KWh.
  • CO2 in the UK is between 0.43 and 0.52 KG/KWh for electricity. An average tree absorbs 14KG/yr.