Observations on dual PSU

Installing three fully stacked Power Edge R430's yesterday I was wiring them into an A and B feed power distribution unit (PDU) that I had just put into the rack. Usually this goes without thought as there will be a number already displayed on it in terms of power draw in Amps (even if this is a little random, and not to be taken as "the truth" - but is a good guide). As I plugged each of the power supplies (PSU's) into the two new PDU's showing zero amps draw, after starting up, one PDU saw the presented draw, the other saw negligible, displaying zero.

In short the Dell units appeared to be using one of the power supplies, with the other as a hot spare.

I had tested an HP Proliant DL380 G3 (real yesteryear 2A draw 2U 6x 15k RPM SCSI drives monster) of my own in the past with a power meter on each feed to find that it essentially balanced the draw between the two supplies which made sense for a number of immediately obvious reasons such as load shared between them, nothing to take a surge when the other one dies - all good.

So... this begs the question why would one manufacturer take one approach, and was one better than the other?

More modern HP's allow the power to be handled over the two power supplies in a number of ways. These profiles are handled within its configuration. However the Dell seems fixed. So - what are the possible operation modes?

Shared Load - the load is split between the two as per the resistance they present - like a batteries in parallel;

Hot Spare - the device is on, and is driving a nominal load at all times;

Cold spare - this is generally a really poor idea, as the other supply HAS to last long enough for this to realise there is an issue and energise. This is generally unrealistic.

Dynamic - now here is the wonder - what this is is not immediately obviously.

So it begs the question - what is the modern unit doing with once it is up and running, just a single PSU running?

It transpires that after a little reading around the answer is simple - EFFICIENCY.

Something I came across with our UPS which became more and more efficient the harder we drove it (within reason - in this case I believe it was something like 80% of its capacity it was the most efficient). Due to the nature of its components (not tremendously different from a PSU - converting power one way and then back again) - there is a range within which it is efficient. PSU's are rarely asked to deliver their full capacity in terms of load. They may be rated at say 750watts - however their efficiency curve means they are happier at point X or Y. However somewhere along that curve they have been designed to be efficient at a given point.

So lets ponder this for a moment - although it is not clear how these figures are returned.


Here is a dull as demonstration video from the nice people at Dell. Their approach appears to be shared, then make a decision and move to sleep or standby situation, then back to shared if it is more efficient to run both at a lower draw. You are balancing the load between the two to keep one or both as close to their efficiency sweetspot as you can. This is the concept of High Efficiency Mode in the curves above.

Curiosity be gone. Nice. So there you go ... who knew? (me now)




Leave a Reply

Your email address will not be published. Required fields are marked *