Fog Bank in a co-locate
Jay Adelson
adelson at equinix.com
Wed Oct 7 18:25:38 UTC 1998
I'll pass these specs along. I appreciate your comments about negotiation
with the colo provider...but seeing as I *am* the colo provider, my
perspective is that there is very little window for negotiation... Either
I provide your spec or I don't. I haven't found it difficult to meet
specifications such as these and enter into SLA for them.
I expect all of you to hit me pretty hard on these sorts of things, so
the more input now the better.
-J
> Donelan's Climate Settings:
> 8,500 BTU/hr heat dissipation per full rack
> Not all of it must be installed on day 1, but I want to
> see pre-planning how the HVAC will be expanded to meet
> the load (e.g. enough room reserved, aisles wide enough,
> to bring in additional units later, pre-plumbed for the
> outside heat exchangers, compressors, etc).
> Minimum 1 air change per hour
> This is actually a 'clue-check' question. How the co-locate
> operator answers it will tell you if they really know their
> HVAC systems. Unoccupied areas can get by with as little
> as 1 air change per 24 hours, but if you are doing an install
> or other work in the area, you'll appreciate more frequent
> air changes.
> Humidity: 35% to 50% R.H., non-condensing, max 2% change per hour
> Ideal: 40%-45% R.H., measured at three locations with max 5%
> variation. Humidity is expensive to control, so it is
> common to have R.H. at the high end of the range (e.g.
> 45% R.H.) in areas naturally more humid (e.g. northeast)
> and R.H. at the low end of the range (e.g. 40% R.H.) in
> areas naturally drier (e.g. southwest).
> Maximum (not to exceed 24 hours/annually): 20%-60% R.H.,
> non-condensing
> Temperature: 59f to 77f (15c to 25c), max 4f (2c) change per hour
> Ideal: 72f (22c) at my rack, but since it is econmically infeasible
> to control the temperature at the rack level, I normally
> ask for 68f (20c) as a room average, measured at three
> different points, with a maximum of 4f(2c) variation.
> The assumption is at worst I'll end up with 72f(22c) at
> my rack. Anything much colder than 68f(20c) is a bit
> harsh on humans that need to work in the area. A tech
> thinking about how cold he is, isn't thinking carefully
> enough about the job he is supposed to be doing.
> Maximum (not to exceed 24 hours/annually): 49f(10c) to 85f(30c)
>
> You'll notice I tend to be metric-centric in my measurements. These are
> much stricter than your typical C.O. environments.
>
> Its more a practical matter of negotiating with the co-locate operator,
> and asking for more than you really need on the assumption even if they
> deliver just 50% of what they promise it won't be an immediate disaster.
> They may have no problem delivering 68f(20c) when the co-locate is 90%
> empty. But when it fills up, I'm betting with the lower the starting point,
> the less likely I'll run into my upper limit.
>
> I choose most of the values to give myself enough of a buffer before
> exceeding any equipment parameters, so I can try to negotiate a resolution
> with the co-locate operator before the levels start to causing equipment
> to shutdown. Otherwise if you set the limits at the maximum your equipment
> will tolerate, a co-locate operator may not do anything until after the
> temperature goes over 104f(40c), but by then its too late. Obviously the
> co-locate operator wants to write the contract with the largest acceptable
> environmental ranges possible, and the co-locate lessee wants the
> narrowest acceptable environmental ranges. Its a barginning point, how
> much do you really want to pay for +- 1 degree, or +- 10 degrees.
>
> In reality, most modern electronic equipment will operate well in a wide
> range of human-habitable enviroments, as long as the enviroment is stable.
> Its the oscillations that will kill your equipment every time. And in most
> cases, if the price is the same, I am willing to trade a particular value
> for tighter tolerances, e.g. 75f+-1f. But the price is rarely the same,
> so I generally choose the bigger range as shown above.
> --
> Sean Donelan, Data Research Associates, Inc, St. Louis, MO
> Affiliation given for identification not representation
>
---
[ Jay Adelson adelson at equinix.com ]
[ Chief Technology Officer Work: +1-650-813-9031 ]
[ EQUINIX, Inc., Palo Alto, CA Fax: +1-650-858-8368 ]
More information about the NANOG
mailing list