A friend of mine told me that the company he made an internship at didn't have server racks, because the cooling just takes up too much energy and thus isn't effective. That company was using "normal" PC Towers to house their servers. Those Towers where placed in some kind of heavy duty shelves.

Is it really the case that cooling a server rack needs more energy than using towers?

  • Thank you very much Brent Pabst. I'm not a native speaker and my english got a little rusty over the last few years. :) – JohnRW Nov 16 '12 at 15:42
  • Arey ou sure these were desktops? Or were they pedestal severs>? – Zapto Nov 16 '12 at 15:43
  • @t1nt1n I think he is talking tower servers as opposed to rackmount. – Brent Pabst Nov 16 '12 at 15:44
  • 2
    strange thinking at best, the only issue that I can see is due to server density with racks being more dense thus concentrating the heat. Just think of having hundreds of towers you'd need a football field to house them but they would probably remain cool enough. Don't think the trade of is a good one though. – tony roth Nov 16 '12 at 15:46

I'd actually argue that the opposite is true-- most rack-mountable machines are designed and purpose-built to be server computers. Today's modern rack-mountable servers are designed with energy efficiency in mind, whereas "tower" computers, while potentially still designed with energy efficiency in mind, aren't typically designed for server workloads.

In a hypothetical scenario with all things being equal between two machines except for the case and mounting hardware basic physics says that the heat output of both machines will be the same. Racks, however, have the advantage of allowing you to construct "hot" and "cold" aisles in the datacenter-- that is, rows of machines with their fans all pointing in the same direction sending their hot exhaust air directly into ducts while pulling cool intake air from the other side of the rack. This isolation typically allows the HVAC system to operate in the most efficient manner. Tower cases typically aren't amenable to this kind of isolation and you end up with intake air being "recycled" exhaust air. Concentrating the heat is a Good Thing(tm) if the heat can be efficiently ducted away from the servers.

If the company only has a few server computers they're not going to have an appreciable difference in server energy usage expense (assuming that the "tower" servers they're using are as energy-efficient as equivalent rack servers-- "build it yourself" white box "tower" server computers will not achieve anything close to the efficiency of a purpose-build rack server computer from a major manufacturer).

It sounds like they didn't have enough machines (or a sufficiently small space) to worry about active cooling and they were just getting by with passive cooling. That's not a function of their using "tower" servers, but rather a function of the size of the room, airflow, ambient air temperature, and the small number of machines outputting heat. With any quantity of servers or a confined space with poor airflow, though, they would need active cooling and, at that point, there are major efficiency gains that can be had by using rack mountable equipment and ducting the exhaust properly.

Of course, cooling and efficiency are only part of the equation. Technician hours are expensive, and nice things like integrated cable management, telescoping rails, cable management arms (I like them, personally), and the general "designed for servicing" nature of quality rack-mountable server gear makes for less time spent when servicing failed components. Try quickly disconnecting a "tower" server from a mass of tangled cables (as I've so frequently found behind the typical "tower servers stacked on wire racks" installations).

  • +1 Beat me to it. Most servers now are pretty darn efficent. – Zapto Nov 16 '12 at 15:43
  • 1
    +1 In addition to being purpose-designed, rack mounted systems (in a proper datacenter, with proper airflow) will often use less energy (spinning fans) to cool themselves. The companies that design these things spend LOTS of money on computational fluid dynamics models of the airflow through the system. The other benefits from easy service are also not to be overlooked. – voretaq7 Nov 16 '12 at 16:37
  • Thanks to everyone who answered. You cleared a lot of things up for me =D – JohnRW Nov 16 '12 at 18:17
  • If your supposition about them largely relying on passive cooling is correct, typical towers do have an advantage in having more room for large heatsinks than rackmount servers where making them as thin as possible to maximize rack density is a major design factor. – Dan Is Fiddling By Firelight Nov 16 '12 at 18:28

I think your friend is a bit mistaken in this statement and given that it was an internship its possible he/she is not very familiar with how cooling in cabinets work.

Regardless of the form factor of the computer/server every case has built-in cooling mechanisms be it fans or some sort of liquid cooling system. Either way when you place a lot of servers (see high density) into a confined space the ambient air has to be cold enough and able to replace the hot air pushed out of the back of the server cases with cold air, otherwise the ambient temperature of the room will slowly rise. This happens regardless of rack chassis or tower chassis.

So, no the cooling requirements are still dependent on the size of the room, the density and many other factors. The only thing that changes is the size of the AC unit required to offset the heat generated.

  • Thanks to everyone who answered. You cleared a lot of things up for me =D – JohnRW Nov 16 '12 at 18:17

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.