Modular Data Centers
Data Centers: Modular vs. Conventional
A self-contained portable data center unit that can be configured to be used to replace, expand, or modify existing data centers, or used instead of a conventional data center.
Item ID: 485
Sector:
Commercial, Industrial
Energy System:
Multiple Energy Systems--Design
Technical Advisory Group: 2013 Information Technology TAG (#8)
Average TAG Rating: 2.38 out of 5
TAG Ranking Date: 10/25/2013
TAG Rating Commentary: - New construction oriented; baseline challenges; persistence or EUL concerns
- These represent the fastest way to drastically impact an existing building DC, drop in an entirely new DC and recover the internal space
- Not an ET for most systems.
Synopsis:
This is a "server racks in a box" solution. Several manufacturers make containers -- often literally a shipping container, and often called a "pod" or module -- pre-loaded with racks of servers. Manufacturers provide pre-loaded modules, or customers can custom-design their own modules. This offers a number of efficiencies. For instance, the server room should be less expensive than the same size of custom data center. They are also pre-designed, typically with very efficient airflow management and cooling strategies. Most of them have a Power Usage Effectiveness (PUE) of 1.2 or less, with some manufacturers claiming as low as 1.04. A module, if it meets the data requirements of the organization, can be used as a self-contained data center, or added to an existing data center to expand or modify the center, and multiple modules can be ganged together to make up a larger data center.
An example of the container-type data center module is the HP POD 240a, which typically comes pre-built with 44 50U racks, for a total "U-space" of 2200 (“U” is a “rack unit,” and refers to a space 1.75” high in a server rack that in some cases could hold a single server).
"Another form of modular data center fits data center equipment into a facility composed of prefabricated components that can be quickly built on site and added to as capacity is needed. For example, HP’s version of this type of modular data center, which it calls Flexible Data Center, is constructed of sheet metal components that are formed into four data center halls linked by a central operating building." (Wikipedia, 2013)
Baseline Example:
Baseline Description: A data center with equivalent computing power to an HP-style shipping container-sized data center module
Baseline Energy Use: 6824 kWh per year per square foot
Comments:
Determining energy usage and a baseline is particularly problematic for modular data centers. Not only is it difficult to estimate what the loading would be on a typical module, but it’s even more challenging to figure out the baseline, or what it might be replacing. A reasonable way to approach this is to figure out a “typical” energy usage of the module, then estimate how much more efficient this might be than what it would be replacing, and “reverse calculate” the baseline.
The estimated energy usage of a module is based on a Hewlett Packard (HP) "POD 240a" fully loaded and operating at an average of 20% of nameplate capacity 24/7 or 8760 hrs. per year. They are 21.5 ft. by 45 ft (968 s.f.), with a “redundant IT capacity” and a power rating of 1152 kW (assume the other 1152 kW of redundant capacity is not operating). The container holds up to 44 50U racks for a total U-space of 2200. This is assuming the module has a Power Usage Effectiveness (PUE) of 1.1 (HP claims 1.05). Thus annual energy usage of the module would be:
1152 kW x 20% load x 1.1 PUE x 8760 hr. / 968 sf = 2294 kWh per year per square foot.
HP designs and builds these modules in quantity, so it is reasonable to assume that the computing energy usage of a typical custom-built data center would be noticeably higher than the HP pod. Let’s assume they use about 50% more power than this module. This PUE is also extraordinarily good. An average data center has a PUE of 2.9. In addition, a typical data center would not be this dense, so it will actually occupy more square feet, but for an apples-to-apples comparison, and to still be able to use the “per square foot” unit, we will assume that the baseline data center fits into a hypothetical space the same size as the module. Furthermore, assume that the server utilization rate is the same for both the baseline and the modular center. With these simplifying assumptions, the baseline energy use would be:
2620 kWh per year per square foot x 150% x 2.9/1.1 = 9070 per year per square foot
Manufacturer's Energy Savings Claims:
Currently no data available.
Best Estimate of Energy Savings:
"Typical" Savings: 60%
Low and High Energy Savings: 30% to 80%
Energy Savings Reliability: 2 - Concept validated
Comments:
The Modular/Container Data Centers Procurement Guide: Optimizing for Energy Efficiency and Quick Deployment, developed by Lawrence Berkeley National Lab, determined the PUE of various types of modular data centers, noting the number of manufacturers of each type. The PUE of a water-cooled chiller plant is roughly the median of the various types, and this type has as many manufacturers as all other types combined. Therefore, this seems like a reasonable value to use for a typical value, which is 1.13. If a typical data center has a PUE of about 2.9, that would be a savings of about 60%. (Lawrence Berkeley National Labotory, 2011 Pg 19)
Energy Use of Emerging Technology:
2,729.6 kWh per square foot per year
What's this?
Energy Use of an Emerging Technology is based upon the following algorithm.
Baseline Energy Use - (Baseline Energy Use * Best Estimate of Energy Savings (either Typical savings OR the high range of savings.))
Technical Potential:
Units: square foot
Currently no data available.
First Cost:
Currently no data available.
Cost Effectiveness:
Simple payback, new construction (years): N/A
Simple payback, retrofit (years): N/A
What's this?
Cost Effectiveness is calculated using baseline energy use, best estimate of typical energy savings, and first cost. It does not account for factors such as impacts on O&M costs (which could be significant if product life is greatly extended) or savings of non-electric fuels such as natural gas. Actual overall cost effectiveness could be significantly different based on these other factors.