RE: DATACENTER: AC and colo sizing question.

[Sean Donelan <sean@donelan.com>]

On Mon, 14 February 2000, "Christopher E. Brown" wrote:
> 
> I would say 65F rather than 75F, with an absmax of 70F.

Most modern electronic equipment can operate in a "nominal"
office environment.  Unfortunately, you rarely find this
environment outside of textbooks.  There is also telco/NEBS
equipment which can handle greater environmental extremes, and
computer room equipment which can handle less environmental
extremes.  But if you have one of the later, you usually know
the specific requirements you have to meet.

You throw it all together, and what do you end up with?
It seems like many people punt, and over-specify their
requirements.  This is very energy ineffecient, and costs
you a lot even in the short run.

I've found 72F (+/- 3F) will satisify nearly 100% of electronic
equipment commonly found in computer rooms.  Even the most
temperamental pieces of electronic equipment have a maximum
recommended temperature no lower than 77F (25C).  Note: all
measurements taken at the equipment air inlet, not at the CRAC.

The extremely few things which require even stricter environmental
controls, tend to come with their own planning guide, dedicated
chillers, and so forth.  The bench around a Cray supercomputer wasn't
just for looks, it was functional.

> Whatever the size of the room good airflow is a must, hot and
> cold pockets are a *very bad thing*.

You point out the most common problem in datacenters, poor airflow.
People try to compensate by lowering the temperature, but this causes
other problems such as humidity control and condensation.  Neither a
steam bath nor a meat locker are good for electronic equipment.

> The equipment is designed to operate with a max intake air
> temp of X with Y CFM flow.  You must maintain an input air temp well
> below the devices rated max, and prevent obstructions to the airflow.
> 
> This means paying attention to where devices exaust to, and
> keeping said exaust from the intake for other devices.

Yep, a rack is its own mini-climate.  The temperature can vary tens of
degrees from the bottom to the top.  Injecting super-cooled air into
the bottom, hoping it won't be heated too much by the time it reaches
the top isn't the best method.  But rack design seems to be a lost
art.  In the "old days" (i.e. 10 years ago) there were specialists at
all the mini and mainframe companies which took great care planning
cabinet layouts.  In today's world of Do-It-Yourself server assembly,
the goal seems to be cram as many ascend maxes into a 84" rack as
possible with no thought.

Patient: Doctor, Doctor, it hurts when I do this.
Doctor: Don't do that.