[greyboy]

How long would some <X> 55-gallon drums full of fuel have helped, anyway? Most cars can easily run through 55 gallons of fuel in a matter of hours (<10), much less a datacenter!

There well may have been poor planning at some point, but these things happen so infrequently there must be an allowance for this-is-a-disaster that cannot easily be worked around.

Another question is for the customers: are they running all their services from a single datacenter? Sounds like it shows poor planning.

[patrickgzill]

It is an easy question to answer, e.g. http://www.pmsi-inc.com/pdf/GeneracPowerSystem.pdf shows 1MW generator uses about 63 gallons per hour, just a little more than 1 drum.

I don't know the size of the facility, so I can't tell how many drums it would take, per hour, to keep the place running.

[greyboy]

So, now they're expected to move around an unknown quantity of 400-500 lb 55-gallon drums of fuel? And where would they have sourced these so easily and quickly (another unknown, I would think)?

We don't know how much energy is being used nor how much fuel is required to run the generator(s) per hour, among other things. That seems like a lot of important information that's missing to call this situation 'poor planning.'

But, I'll respond to your single datapoint with mine: according to [1] that's 2 full drums for every one hour of generator running.

My point isn't to be simple argumentative but to look at things from a more appropriate perspective. Generators in these circumstances (and my professional experience) are not meant for very long periods or indefinite usage.

I'm filing this proposed drum-filling plan into the 'unrealistic' category.

1) http://datacentersmadesimple.com/tech_highlights.html

[patrickgzill]

Try to see it from a different perspective: customers are paying (my guess) $800 per rack, per month, plus power and bandwidth charges; and a rack takes up 30 square feet once you include the space around that rack.

So almost $30 per square foot for just the raw space alone. $360 per square foot per year is a high rent, even for NYC.

And what is the client supposed to get for his money? Reliability! The engineering and facilities management expertise to ensure this, is baked into the costs.

You ask, "so now they are expected to move 55 gallon drums of fuel"? ABSOLUTELY they are expected to do that. The only "appropriate perspective" is that the clients are paying a lot of money for the datacenter to do whatever needs to be done.

They had a week of warning to source these; they already have a long-standing relationship with their fuel supplier for diesel delivery, so they call him up and say "Joe, we need 20 drums of diesel in addition to topping up the tanks we have" and they arrive in the next 2 days.

These diesel generators are basically modified / tuned versions of a big truck or marine diesel, which has a rebuild interval of 500K to 1 million miles if used as a truck engine or some high number of operating hours (like 10,000 hours). Perhaps you are thinking of LPG, natgas or gasoline powered gensets, which are designed for less frequent use.

I researched all aspects of building a DC years ago and realized that even if I could raise the $5 million to do an entry level one, my effort was best spent elsewhere.

Customers punish downtime, this DC will lose clients, be sure of it.

Aside: there was a guy in New Orleans who kept his DC running all through Hurricane Katrina and after it - if you search the site at http://mgno.com with terms like "diesel drums" you will find his old posts. Can't seem to easily link to these old posts, though.

[greyboy]

I'm well aware of these generators with two shipping-container sized units right outside my building, tested fequently. So, I concede and simply disagree.

I agree they will likely lose some customers but I disagree that there was too much they can do now. Were they in the mandatory evacuation zone? (I don't know) Will 20 drums (10-ish hours) really help if this is a multi-day outage? Did the customers plan for a failover to another datacenter, or put all their eggs in one basket? (oops!)