Every once in a while I’m dragged into a data center power discussion, especially if I’ve prepared a big configuration or if the system needs to be placed in a data center that is pretty stressed power-wise. In most of those conversations it’s only a matter of time until someone starts mixing up the units: kW vs kWh. Since they both mean something completely different, it’s important to use the correct one.
Let’s dive right in. When IT people talk about power they usually intend to use kW (kilowatt). kW is a unit of power. Unfortunately they often use kWh (kilowatt hour) which is a unit of energy. Let’s apply this on an analogy that most IT guys can relate to: cars. Power is measured in bhp or kW, energy is measured in…. gallons of fuel that can fit in the fuel tank!
So if you are one of those people that mixes up the units: Congratulations, you’ve just sent me an email that is the equivalent of telling me your flashy red sports car has 60 liters of gasoline worth of energy. Exciting!
Watt! What is a Watt?
Watt is a unit of power. In a heating appliance it describes the amount of energy it can convert (or burn) per second and thus how rapidly you can heat a room. IT equipment needs a certain amount of power to function otherwise it’ll do nothing but look pretty.
The fact that I’ve used “energy per second” should make you curious and think: wait, so I can break up a Watt in even smaller units, right? Correct! 1 W is the equivalent of one joule per second (or in short J/s).
Kilo (lowercase k) means a thousand, so 1 kW means 1000 W which is in turn 1000 J/s (joules per second). Take your heating appliance and assume 100% efficiency: if I feed it 1 kW of electricity it will generate 1000 joules of thermal energy per second to heat the air in my house. If I feed my SAS drive 15 W of power the little motor will spin the platters with the required amount of rpm, move the heads across the platters, power the PCB and all the chips on there to read/write the data and generate heat as a byproduct (friction of the spinning platters, resistance in the PCB and chips, etc). Or maybe you’ve got an SSD drive installed which does not need to spin a couple of heavy platters and will do just fine with only 5 W…
Energy and easy of notation
So let’s assume you’ve got a rack of IT equipment. You’ve filled it to the brim with storage equipment. You’ve used high density enclosures so the entire rack consumes 20 kW of power, non-stop, all the time. How much energy does it use hourly?
Easy: so we’ve got 20 kW. That’s 20,000 W, which is 20,000 J/s.
An hour has 3600 seconds (60 seconds in a minute, 60 minutes in an hour).
In an hour we will use up 20,000 J/s * 3600 s = 72,000,000 joules. A lot of zeros is annoying so let’s shorten that to 72 mega joules or 72 MJ. If we kept that rack running for a day we’d use 72MJ/hr * 24 hours = 1728 MJ or 1.728 GJ.
So I think it’s obvious that this is… less than ideal. If you use a lot of power or measure over an extended timespan you’ll end up with numbers that are bizarre in size and pronunciation. Not ideal for the average consumer. So let’s make it a bit easier and make up a new unit: kWh!
kWh means kiloWatt-hour. Writing it down as you’ve learned in your physics lessons on high school it would look like: kW*h or kW times an hour. In the real world it’s the amount of energy that you would use if you used 1 kW of power for 1 hour. Back to our rack: powering 20 kW for 1 hour uses 20 kWh of energy. Easy right? If we kept the rack online for 24 hours it’s 20*24=480kWh. One year later we’ve used 175,200 kWh or close to 175 MWh. Much easier to calculate, right?!
The observant, scientific, math oriented reader now thinks: hold on, you’ve got two units of time in there. And how many joules fit in a kWh?
1 kW is 1000 J/s. And kWh multiplies that with an hour which is 3600 seconds. So effectively 1 kWh is 1000 joules per s TIMES 3600 seconds. Which is 3,600,000 joules or 3.6 GJ.
kW vs kWh
Another analogy. Energy = amount of sand. Power = kilograms of sand I can carry around per hour in my wheelbarrow. If I need to move 1000 kg of sand in steps of 50 kg per wheelbarrow, I will have to move 20 wheelbarrows. I can either move 20 wheelbarrows in an hour and be done with it, or I can slow to a crawl, push one of them per hour and be stuck with the tedious job for 20 hours.
The same principle applies to power. A rack that requires 20 kW of power will use 20 kWh of energy in an hour. A second rack that requires only 10 kW of power can run on the same amount of energy for double the time.
Power and energy in IT
So what does this mean for IT in real life? Where will you run into these units?
- Cabling and fuses: you can only run a finite current through a cable before it melts. A typical data center PDU (Power Distribution Unit) fuse is rated at 32 Amps. Multiply this with the voltage that your data center runs on (240 Volts over here) and you’ve got the max amount of power you can put on your PDU: 32 Amps * 240 Volts = 7680 W = 7,6 kW. Install four of these PDUs in a rack, only load them for 50% to cover for outage of a feed and your rack can sustain 15 kW of power.
- Cooling: if you use a lot of power you will generate a lot of heat. Energy never disappears but is only converted: thermal energy is usually the end-state. Stuff 40 kW of equipment in a rack that’s only capable of extracting the heat at a 20 kW rate and you’re sure to experience a mini meltdown in your data center.
- Billing: for a domestic power connection the energy provider is only interested in the amount of energy you use, not really in what manner you use it. You will pay for the amount of energy or kWh, not for the power (KW).
Commercial/industrial connections are a different story and are often also charged based on their peak power load because they have a much bigger impact on the energy provider. If your factory quickly alternates between 10MW and 0MW, the power company will have to employ a DJ to constantly push a little slider that increases/decreases production… So they want to motivate you to level off at 5MW non-stop.. saves them one paycheck 😉
“You know what I mean!”
Yes I do. And silently laughing on the inside because you’ve asked me for a rack that “has 10 kWh“ and I’m picturing myself unplugging your rack once you’ve used all your allotted energy is still funny. But use the correct units. It’s not that difficult.
It’s easy to say that IT equipment is on 24/7 and uses the same amount of power all the time. But this hasn’t been true for quite a while now: processors scale back depending on load, drives stop spinning, VMware DPM switches off hosts depending on the cluster load.
Or the other way around: backup is usually your most demanding application and can actually increase your power consumption at night due to deduplication eating up CPU cycles and your disks swapping non-stop.
And I’ll keep challenging you regardless.
Purely in the interest of fire safety of course! Because I never know when I’m going to run into someone that knows the difference, asks me for a rack that can deliver 40 kWh and then secretly shoves 400 kW of heaters in there and switches it on for 6 minutes (after disabling the fuses).
Questions, comments, feedback?! Feel free!