In terms of Reliability...

[PQexplorer]

In terms of Reliability, Which of these UPS parallel(redundant) configurations would you find more reliable for a data center that needs a total of 160kVA available power?

1+1 (a pair of 160kVA).
2+1 (Three UPS of 80kVA).

IMO, with a poor mathematics intuition. the first config. is more reliable, because you have a 50% of chances to lose redundancy, while in the second one you have 67%....Isn't it?

 

[steve66]

Re: In terms of Reliability...

The first is more reliable. To have an failure, both UPS's would have to go down.

On the second one, you would still need 2 UPS's to go down for a failure. But that is more likely since there are now 3 UPS's.

But the question doesn't seem fair. The first option has 320KVA of UPS, but the second only has 240KVA.

I would consider (4) 80KVA UPS's (that gives the same KVA as the first option), and would give a very high reliability.

Of course, I am ignoring all the switching and equipment needed to make UPS's run in parallel. I'm don't really know a lot about how they parallel UPS's.
Steve

[ September 15, 2005, 01:07 PM: Message edited by: steve66 ]

 

[charlie b]

Re: In terms of Reliability...

Now you are kicking up some of the dust in the attic of my mind. My courses on reliability theory are over a decade old, and my text books are at home (not with me now).

I do recall that determining the probability of "system failure" has to start with defining "system failure." I presume here that a "system failure" means that the available supply of power is any value lower than 160 KVA. In other words, you call it a "system failure," even if you can supply some loads, since you need to be able to supply all loads.

Two other factors in the equation for probability are the probability of a "component failure" (i.e., loss of any single component), and the number of components that must fail before the "system as a whole" fails (i.e., the level of redundancy).

Let us assume that the probability that a single 160 KVA unit might fail is the same the probability that a single 80 KVA unit might fail. In both of your cases, it takes two components to fail, before you get a "system failure." My instincts tell me that they two options have precisely the same reliability. But I would need to read old textbooks, to be sure.

I agree with Steve that four units of 80 KVA each would be, by far, a more reliable system. The simply reason is that it would take three "component failures" to get a "system failure." I think that the fact that this would give you the same total capacity is not relevant to the question of reliability. The important question would be, "Is it worth the extra cost to buy a fourth unit?"

HOWEVER. If you were to tell us that there is some value to your company to be able to supply a portion of the loads, if having 80 KVA of capacity is better than having 0 capacity, if having 80 KVA is not a "system failure," then the problem gets far more complicated.

 

[rcwilson]

Re: In terms of Reliability...

One axiom I was told about reliability is the more parts there are, the greater the chance for a failure.

I agree with Charlie's analysis about the probability of a failure being about the same for the different UPS's. But my gut feel is that the simpler, two UPS system would be more reliable because it has fewer breakers, transfer switches, bypass switches, synchronizers, chargers and controls to go wrong.

My experience (some of it maintaining large UPS systems under contract) is that the auxiliary equipment creates more problems: batteries, the in and out switchgear, the panelboards and PDU's fed by the UPS systems. Failures in the downstream equipment kill the loads.

Also consider whether you want 1, 2, 3, or 4 battery systems. A single battery bank with two parallel strings is a common approach to minimize costs without a major impact on reliability in multiple UPS installations.

I like Charlie's idea of knowing your loads to decide if having just 80 kW for the really critical loads would be any benefit.