• ef•fec•tive –adjective
  1. adequate to accomplish a purpose; producing the intended or expected result: effective teaching methods; effective steps toward peace.
  [Origin: 1350–1400; ME < L effectīvus practical, equiv. to effect(us), ptp. of efficere
• ef•fi•cien•cy –noun, plural -cies.
  3. the ratio of the work done or energy developed by a machine, engine, etc., to the energy supplied to it, usually expressed as a percentage.
  [Origin: 1585–95; < L efficientia, equiv. to efficient- (see EFFICIENT) + -ia -Y3 ]

The first thing we notice is that effective is an adjective and efficiency is a noun. That tells us something important. The upshot of the difference is that Efficiency is a thing – something that can be expressed in finite terms with or without modification from the use of an adjective like Effective, which does not exist in and of itself. Let’s explore this some more in practical terms.

Cars are usually a good starting point to evaluate the use of data center energy terms, as are teenagers and the household environment but we wont go there right now, since most of us can relate in real-terms to the practical side of automobiles and/or motorcycles/scooters/bicycles if so inclined.

In examining “effective” consider that driving a car around a shopping center for an hour while trying to grab the prime parking spot in the front row by the entrance is not an effective use of the vehicle, time, or whatever precious resource being used to move the car – even if your driving a Fred Flintstone mobile! A more effective practice in the fine art of parking at a shopping center is to either

1. make a quick pass by the coveted spot then grab anything nearby,
2. forget about trying for the spot as you know someone is camped there all day so you take the first spot you find,
3. park out on the edges of the lot or adjacent to some sort of barrier to protect your vehicle from potential door-dings then walk it in the rest of the way, or
4. get creative and annoy everyone.

Both 2 and 3 will get you out of the car the fastest and on your way into the shopping center though potentially with a slightly longer walk easily justified as additional exercise or reduced CO2 emissions (oh sure!). No comment on #4 – you know who you are.

“Efficiency” comes down to the amount of useful work per unit of energy, which in this case would be miles per gallon / km per liter / liters per 100 km at least with regards to the motor vehicle being used to transport one to and around the shopping center in search of a parking spot. So in this case a hybrid like a Toyota Prius would more efficient than an SUV or truck like a Ford F150. Especially if you insist on being in-effective and drive around the parking lot in search of the perfect parking spot. Note: I do not drive a hybrid or a pick-up and both are fine for their intended purpose, just like most data centers when originally built.

So let’s get back to data centers, PUE note: PDF (power utilization effectiveness), and efficiency. Given that effectiveness and efficiency are not one and the same, what is all this PUE stuff about really? Well I’ll kick this off by suggesting that, as defined within the Green Grid’s document, this is nothing more than a measure of the ratio of where data center power flows. It has little to do with either effectiveness of efficiency as commonly defined.

The PUE metric is clearly a ratio that expresses the relevant proportion of your purported IT power consumption (more in a future blog) to your assumed total data center power consumption by dividing the total data center power by the IT power. Total DC Power / IT Power = a number greater than 1 or it can be expressed as something like 5:3. That clearly looks like a ratio to me. Nowhere in the fundamental equation do we see anything that one can either relate to effectiveness or efficiency as we understand these concepts today.

If PUE was in and of itself about effectiveness, we could, by understanding the ratio of power delivery within the data center, make some intelligent decisions regarding how we are using it. Such as how well it is processing our data, how well it is performing in that data processing (or storage or streaming media or network capacity) task, or even something as simple as how well it is doing today versus yesterday. Now if we had a model for our specific data center’s PUE over all potential operating conditions, temperatures (both internal and external), and a host of other key parameters, then we may be on to something. But without significantly more data, we wouldn’t have very much to work from to know what to tweak to “improve” performance as PUE says nothing about output or useful work.

Going back to the car model, we can develop concepts on how to be more effective in the use of the car/truck/motorcycle/scooter/bicycle based upon the intended use at the time – such as in searching for a parking spot. Knowing the EPA rated fuel economy rating for our vehicle (rated in Miles per Gallon – a form that expresses useful work) we can make some intelligent choices on how we operate our vehicle in search of the perfect parking space.

So what about data centers? Simply put, unless you know the MPG equivalent for your data center, attempting to make decisions on PUE alone as you rev your servers, storage, and network devices in search of bits to crunch, file, or stream you are just wasting your time. Because trying to drive less power to the “infrastructure” side may well be an attempt to repeal the laws of thermodynamics while reducing IT power, though clearly beneficial in terms of reduced expenses and CO2 emissions, will not change your PUE a single bit.

What is the answer? First, don’t confuse effectiveness with efficiency especially when speaking about complex systems such as data centers. Then, let’s focus our efforts on defining MPG for a data center AND creating ENERGY STAR™ ratings for IT devices. You do not buy a car on top-speed alone (especially given both the cost of fuel and the financial/social penalties for speeding) so you should not make that your single criteria in selecting IT hardware.

For a related look at the entire MPG for Data Centers concept read the recently published White Paper. We’ll examine the CUPS (Compute Units Per Second) proxy in future blog posts.