FOCUS CLOUD COOLING
Issue 10, June/July
EASY AS PROVISIONING A VM Open questions remain on cooling infrastructure design for facilities supporting cloud computing
hosts clouds pulls engineers out of their comfort zones.
A
There are a number of challenges that make cooling a facility that supports one or several environments configured to deliver compute capacity in a cloud-like manner different from cooling a traditional environment. Such environments are characterized by high degree of virtualization, high density and wide and frequent load fluctuations. They are also usually deployed along with traditional, largely static environments.
Cooling experts agree that a facility housing cloud-computing systems needs a cooling system whose capacity can be quickly adjusted. A highly-automated, variable-capacity cooling set-up however is not the final answer.
MIXED DENSITIES ARE ONLY THE BEGINNING
“The old data centers, for years and years, used the same approach which was an ‘open ballroom,’” said Ken Cooper, a principal of the Silicon Valley mechanical and electrical engineering firm Cooper Preuit. While the advent of blade servers created additional challenges that made the traditional approach problematic – creating hot spots and creating a need for more air handlers and deeper raised floors – virtualization, which is at the core of the cloud delivery model, raises load densities even further.
One answer to dealing with high densities, which in a cloud environment can reach about 10 kW per rack, has been cold-aisle containment.
Density is only part of the problem with cooling a data center that supports clouds and containment is not necessarily the best way to tackle the problem. The biggest challenge is dealing with load variability from hour to hour,
32
www.datacenterdynamics.com
s is the case with almost every aspect of implementing cloud computing in any of its forms, cooling a data center that
as well as from location to location within the data center.
“Most data centers that we do may have some cloud computing (systems) in it but it’s a small fraction of the servers that are in the space,” Cooper said. While load variability from location to location is something operators have learned to deal with, especially commercial colocation and managed services providers, load fluctuations over time are a relatively new challenge.
High-density racks interspersed around the data center can be particularly challenging for cooling infrastructure also because “pressure dynamics beneath a raised floor can inhibit proper airflow patterns,” said Coy Stine, director of data center services at Bluestone Energy Services, a Norwell, Massachusetts- based engineering and project development firm. “Either high-density racks are supplied with too little volume, creating hot spots, or low-density racks are supplied with too much air, creating inefficient mixing.”
Traditionally, operators would overcool the entire room to supply enough cold air to the high-density racks, Stine said. This approach defeats one of the key benefits cloud computing promises: energy efficiency.
Another common approach is adding or taking away perforated tiles or installing tiles of varying degrees of perforation, depending on need for capacity, in any given spot. This approach may not be quick enough for a highly dynamic cloud environment, however.
DYNAMIC AND GRANULAR COOLING CAPACITY IS KEY
Brian Hans, mechanical engineer and senior associate with Mazzetti Nash Lipsey Burch, says a facility housing cloud-computing environments needs to “have an HVAC system that can react to the changing dynamics of the cabinet.” Mazzeti Nash Lipsey Burch is a San Francisco-based consulting and design firm.
In a traditional environment, changing cooling capacity requires manual manipulation of the HVAC system. The speed with which a cloud system’s load fluctuates, however, does not afford enough time for such manual manipulation.
“Changes (in capacity of a traditional cooling system) don’t happen at IT-type speeds,” Hans said. “They happen in dinosaur-type speeds. It’s the difference in being able to have an adjustable system in your data center versus having a static system.”
Engineers have chosen different ways to add dynamism to cooling systems. In a project Cooper Preuit is doing for HP in Colorado, the design is using a vent tile with a damper in front of a rack. The damper is used to adjust the amount of air coming through the tile, according to data coming from about five temperature sensors distributed throughout the height of the rack, monitoring temperature constantly. Some engineers choose racks that have cooling fans at the top, whose cooling capacity is also adjusted in accordance with rack temperature at any given moment.
A CONCERN BIGGER THAN FICKLE RACK LOAD
While there are solutions for quick automated cooling-capacity control at the rack level, large fluctuations in load may create problems away from the computer floor and in a part of the infrastructure where errors can lead to much more dire consequences: the main chiller plant.
“We would typically incorporate some sort of automation that would respond to loads in real time but with the expectation that the load variation is small,” said Vali Sorell, associate partner for critical facilities at Syska Hennessy, about traditional cooling systems.
Today’s automation systems, however, can only go so far in addressing wide and frequent load fluctuations. Even the traditional data center commissioning process is not designed
COOLING A CLOUD NOT AS
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54 |
Page 55 |
Page 56 |
Page 57 |
Page 58 |
Page 59 |
Page 60 |
Page 61 |
Page 62 |
Page 63 |
Page 64 |
Page 65 |
Page 66 |
Page 67 |
Page 68 |
Page 69 |
Page 70 |
Page 71 |
Page 72 |
Page 73 |
Page 74 |
Page 75 |
Page 76 |
Page 77 |
Page 78 |
Page 79 |
Page 80 |
Page 81 |
Page 82 |
Page 83 |
Page 84