Properly sizing an uninterruptible power supply (UPS) is always a bit of a challenge. This is typically because the power specifications from IT equipment manufacturers and the capacity ratings from UPS manufacturers can be somewhat difficult to understand and match.
True power load
In order to understand how big a UPS you need, you first have to figure out the IT power load it will support. However, compiling an IT system's power load is not that simple. It would seem logical to add up the advertised wattage rating posted on the system's back plate for all IT components that the uninterruptible power supply will power and size it based on that total, but reality differs. Power requirements or consumption ratings for IT systems cannot be calculated like other electrical devices. For example, the power consumption for a 100 watt light bulb is pretty straightforward; when it's on, it consumes 100% of its rating (100 watts/hr), and when it's off, it consumes nothing. IT systems are different in the sense that they consume more or less based on demand. The harder systems work, the more power they consume. Unlike light bulbs, however, we never expect computers to work at 100% of their capacity, nor would we want to. In fact, we typically think about upgrading IT systems way before reaching 100% utilization. So the question becomes; why use 100% of the power rating to size a UPS?
We typically use 70% of the maximum power load for all the systems as the basis to size an uninterruptible power supply. That is very generous considering this means the processors, memory, PCI Cards, NICs, I/O on all disks and the fans, etc. are all running at 70% utilization on all systems at the same time. When does that ever happen? If it did, we would already be looking to upgrade many of those systems, as mentioned before.
For example, if we have a number of systems that add up to 50,000 watts (50kW) based on the manufacturer specifications (100%), this means they will draw around 35,000 watts (35kW) at 70% utilization. That would be the value we use to size the UPS. That's a 15,000 watt difference, which is pretty significant.
Leave room for growth when choosing an uninterruptible power supply
Because IT environments are constantly in growth mode, we must factor in some room for future capacity. Once again, working from the maximum system rating instead of a more realistic value can significantly skew the capacity planning results. Using the values provided in our example, 20% growth on the full 50kW is 10kW rather than7kW if growth were calculated on the more realistic 35kW power utilization.
Appropriately sizing infrastructure components is becoming common practice. Many IT organizations can no longer afford to buy more capacity than they actually need, and prefer a more modular approach. Going back to our example, we end up with an estimated capacity of 42kW using realistic values including room for growth, which is still under the initial 50kW if we use full load power ratings.
Volt-Amp or kilowatt?
To make things a little more complicated, most vendors rate their UPS devices in VA, or Volt-Amp, where Volts are multiplied by amperes instead of the more intuitive watts or kilowatt (kW) rating. VA is typically used to size wiring and breakers, whereas watts is the power an appliance could actually draw. Essentially, it is the difference between "apparent power" and "real power" drawn by equipment. To use a simple analogy, the cord for a particular appliance may be rated for 115V and 15 Amps (1725 VA), but it does not mean the appliance itself will actually draw 1725 watts.
Many UPS vendors use a ratio of approximately 90% to convert VA to watts (100 VA = 90 watts). This is usually referred to as the "power factor." Using our earlier example, a 47kVA UPS would power systems that added up to 35kW of IT load (or 50kW at 100% load), and allow for up to 20% of capacity growth (42kW).
Why worry about proper sizing when choosing an uninterruptible power supply?
It may seem like a wasted exercise to try to size a UPS. Why can't we make sure we provide enough capacity to meet current and future requirements? One issue comes with the circuit and cooling requirements for a UPS device. You cannot install a UPS that exceeds your current requirements without ensuring that the electrical circuit to which it is hooked up is matched. Then, you have to ensure that there is sufficient cooling available to meet the heat rejection of the UPS and the load it can power. It may seem like a long way down the road before you need the extra cooling, but it is surprising how quickly this is forgotten and you end up with a cooling problem.
Power efficiency is also something definitely worth considering. Efficiency is typically measured at 90%-plus capacity, so an oversized UPS will have a tendency to underperform and be less efficient than a properly sized UPS.
Battery runtime is another area where oversizing is frequent. Most people have a tendency to want to keep systems up as long as possible in the event of a power failure. However, due to space, weight, cost of acquisition, and eventually the cost of replacement, it's not necessarily a good idea to buy more battery capacity. Batteries also have a limited life span, so why pay for batteries five years from now when you didn't need them for the first three or four years?
The reality is this: If you need to keep systems up during a power failure, you need an emergency power generator, not just an uninterruptible power supply. You have to consider what else needs to be powered for the systems to be useful. In many cases, if the office and user workstations have no power, or if the server room air conditioning is down, what good is it to have enough UPS battery capacity to keep servers up for 45 minutes to 60 minutes? Most UPS vendors provide software that enables the automated and graceful shutdown of systems following a power failure. The batteries' runtime should therefore be long enough to support that function.
Overall, you must take the time to think this through and calculate the right capacity for your environment. Sizing a UPS cannot be handled like adding more memory or disk space to a server in case you ever need it; there are considerations that can drive other costs unnecessarily high if proper sizing is not applied.
Pierre Dorion is the data center practice director and a senior consultant with Long View Systems Inc. in Phoenix, Ariz., specializing in the areas of business continuity and DR planning services and corporate data protection.
Do you have comments on this tip? Let us know. Please let others know how useful this tip was via the rating scale below.
Do you know a helpful storage tip, timesaver or workaround? Email the editors to talk about writing for SearchDisasterRecovery.com.