Posts Tagged don’s corner

Don’s Corner: In-row or Perimeter Cooling, is there really a difference, and if there is, which is better?

Posted by on October 4, 2012  |  No Comments

Don Melchert, Critical Facility Specialist

I love two-part questions, don’t you?  It shows me that the person asking it probably has some experience with the subject and is asking a loaded question.  Usually, the inquiry stems from an experience that didn’t go all too well, or, it went much better than planned and the question is asked more from a validation standpoint than an argument.  I’d much rather prefer the latter, of course, but in data centers, that’s usually not the case, is it?

With any subject that’s controversial, you’ll never win everyone over.  In the argument over In-row or Perimeter Cooling, the subject is more volatile, especially if the inquisitor’s job is to maintain an antiquated (read dinosaur) perimeter cooling system.  You just know they’re thinking to themselves, “If that system goes away…now what?”  Ever see the latest version of Willie Wonka and The Chocolate Factory?  When the machines took over production, Charlie’s dad was laid off, but in the end, he came back to work as a repair technician on the machine that replaced him.  What made Mr. Wonka decide to replace a human with a machine?  Was it production, quality, or floor space?  If you’re a Co-lo provider, floor space is high-dollar real estate!  So, for this article, let’s use floor space as our filter, and to keep the playing field even, we’ll assume a chilled water system.

After some research, and I’m putting that lightly, I’ve found the Liebert DS to be the most commonly deployed air conditioning system for data center use.  Really, I’ve seen them all over the globe!  Why shouldn’t they be the most common?  They’re easy to train on, so you can get training almost anywhere; they have a long service history, so MTTF rates are pretty accurate; they’re supported by a ton of people, so the MTTR rates are well documented, too.  Of course they have to be shut down to make most repairs, but that topic’s for another discussion.  Even though they’re the most common, the Liebert DS, and perimeter cooling in general, is losing ground to more efficient, close-coupled, In-Row and overhead solutions.

Floor space is the name of the game, so what kind of real estate does the most commonly found perimeter unit gobble up?  Simple answer: 21sq ft.  I said “simple answer” because I made a comment above about having to turn them off for repairs.  Which is why I found that most perimeter cooling believers had more than one Liebert in the room.  Even without a calculator, that’s 42 sq ft, folks, and that’s only at N+1 redundancy.  Within that 21 sq ft area, a Liebert DS can cool up to 42kW of critical load.  Not too shabby, but if you have any kind of redundancy requirement, you’re trying to get the most out of that space.  With a Co-location facility, now you’re looking at how many customers can I not fit into the room because you’re forced to give up 21 sq ft of real estate?  That kind of talk doesn’t fly too well in a boardroom.  So what’s the alternative?  Let’s look at one, the APC In-row, close coupled cooling, or IRRC.

The In-row RC, takes up only about 3.5 sq ft, but it makes up for it’s diminutive nature by removing 18kW of heat load.  With reference to floor space (…and this is where, somewhere out there, my high school geometry teacher is laughing her evil laugh), you’d need a bunch of IRRCs to equal the same amount of real estate consumed by a Liebert DS.  The answer is 6, Ms. Davidson (flashback to 9th grade).  With 6 IRRCs, one could cool up to 108kW of critical load!   Before you ask, no, you won’t need to make them redundant, as the most common failure items, the fans and power supplies, are all hot-swappable.  Here’s something else to chew on: the IRRCs don’t have to be right next to each other to get the job done.  At about half a rack wide, they can be separated around the room as needed, but you’ll need a sawzall and a darn good explanation afterward if you try to do the same thing with a legacy style perimeter unit!  Please, if you do this to a perimeter unit, invite me out and I’ll buy lunch?

Knowing that I’d have to prove my math, I worked the number backward and concluded that with 108kW of critical load, one would need 2.5 Liebert DS units to remove that much heat, or step up to the next size unit.

That’s where it really strikes home for me, personally.  After performing a CFEP, I’m often challenged with explaining why its important to let go of the ancient ways and consider converting from a perimeter cooled space to a modern, close-coupled solution.  If floor space is the driving factor, then the math, for me, drives the point home fairly well.  If a Co-lo data center can increase their customer base from 12 to 108kW, simply by becoming more efficient in their cooling architecture, maybe Charlie Bucket’s dad might not be laid-off after all.  In today’s job market, that means a lot.

Don Melchert

Don’s Corner: Who’s Responsible For The UPS, the IT Staff or Facilities Maintenance?

Posted by on September 5, 2012  |  No Comments

Don Melchert, Critical Facility Specialist

Who’s responsible for the UPS, the IT Staff or Facilities Maintenance?

I have been asked this question many times, for all sorts of infrastructure equipment.  From the structure windows to the CRAC, I’ve heard this same question asked both before and after an expensive catastrophe has brought everyone into the boardroom.  To properly entertain answering the question, we have to approach it from two fronts: Old School and New School.  Without first defining those two parameters, there’s no way to stop the finger pointing if it all goes wrong, so let’s start there.

The “Old School” Platform

In times long ago (now if that doesn’t sound like the beginning of a new Tolkien novel, I don’t know what does) anything and everything that did not physically house the critical network was placed under the governorship of Facilities Maintenance.  Quite literally, the only exclusions were the racks, servers, telecom equipment and associated cords and peripherals.  If one wanted to relocate a server to another side of the room, powered from a different breaker, the first mission was to request approval from a Facility Manager.  Wait a second! There still are data centers out there that live under the iron rule of Facilities Maintenance, but that topic is best kept in the file labeled “Future Don’s Corner Topics”, at least for now.

If I were to provide an answer to the “Who’s responsible for the UPS?” question, based on the Old School platform, the answer is:  Facilities Maintenance, with operational approval from the IT Director. The reason for this is, originally, IT personnel focused on the devices and the network itself, and so did the educational institutions that certified the personnel of yesteryear.  This wasn’t necessarily a bad thing.  Consider this; imagine asking a Facilities Engineer in 1987 that works in the sun, rain, wind, and snow maintaining the building’s HVAC system to log into the network.  Just speaking that way to a member of the facilities staff would have caused them to panic, possibly invoking memories of scenes from the movie War Games!  However, this works both ways.  Can you imagine asking a member of the IT staff to check the run hours on the CRAC unit?  They probably would have called security on you, knowing you must have lost your mind.  Few IT personnel would have considered dirtying their hands, and no self-respecting member of the facilities staff had the patience for “C://”.   In my answer, I said “with operational approval from the IT Director.”  If you’re not immediately in agreement, give UNS a call and we’d be happy to help bring you safely back from cryo-stasis and into the new world of today’s IT management!

The “New School” Platform

IT has fought hard for the past 30+ years to improve every aspect of network operations.  This is fact, as I have, willingly enough, seen this from both sides of the fence, first, as a Facility Manager, later as part of the IT consulting community.  Unknowingly, though, we were all working toward the same goal of efficient operations, more with less.  The UPS, the In-Row Air Conditioner, the servers, even the programs we were running…everything associated with the critical IT network became self-serviceable or hot-swappable in some way.   What used to require a shutdown of the network during a midnight to 7am all-nighter to swap out a bad component now requires little or no notification, no shutdown at all, and it all takes place in about five minutes, including unwrapping it from the box!

However, all of this new technology means two things: First, it’s easier for us to access the critical network, so we just expect it to be there waiting for us, whenever we open our iPhones; Second, all of our devices expect the exact same thing, only just like The Matrix, our numbers pale in comparison to theirs, and so does our priority levels on the network.  We, the custodians of our NCPI and its assets, must maintain this uptime requirement and now, more than ever before, the IT community is prepared, both technically and internally, to do so.

If I were to provide an answer to the “Who’s responsible for the UPS?” question, based on the New School platform, the answer is: IT, with operational approval from the Director of Facilities Maintenance.  Together, with sound communication, both are providing a service to the company’s critical network and should be able to count on each other to continually provide primary and uninterrupted back-up power to the company’s Network Critical Physical Infrastructure.  Again, if this does not make immediate sense, please, don’t hesitate to call UNS for an explanation of why this shared responsibility is so important.  The job you save may be your own!

Don Melchert