It’s been a while since I’ve blogged about one of our TradeOff Tools. But as promised, I want to continue to give some insight into what’s behind our TradeOff Tools, so you have full confidence in using them to aid in your data center planning decisions. Today, I want to share some background on the Virtualization Energy Cost Calculator.
This calculator enables you to see the impact that server virtualization and data center design choices can have on the energy and space savings of the data center. The tool shows you the before and after power usage effectiveness (PUE), a breakdown of where the energy savings comes from (i.e. server reduction, power & cooling systems), an estimate of electric bill savings, and the amount of space saved through the project.
Here’s how it works. The total power consumed by servers and the average server power draw is estimated based on three inputs; total IT load, the percent of load that is servers, and the number of servers. Then, based on post-consolidation inputs including the percent of servers that can be virtualized and the consolidation ratio, the tool estimates the new server power consumption. We approximate this power draw with a curve fit of data on power draw as a function of the # of VMs on the machine. Our curve fit formula is: Power draw per server * # of VMs on the server ^ 0.38837. This model was derived from a few sources, including SPEC.org power benchmark, a Sine Nomine Associates paper titled “A Comparison of HP BladeSystem with Thermal Logic Technologies to Competitive Systems”, and a Google paper titled “Power Provisioning for a Warehouse-sized Computer”.
Server consolidation is often where the energy discussion ends for virtualization projects, but this tool includes another critical aspect – the impact of the physical infrastructure on overall performance after a virtualization project. We know that if you leave the power and cooling systems alone, and decrease your IT load through consolidation, your overall energy bill goes down (a good thing), but from the perspective of PUE or physical infrastructure efficiency, the data center just got worse. This tool helps you see the impact of this, and allows you to see how power and cooling improvements like right-sizing the CRAC/CRAH or UPS, replacing outdated equipment with high efficiency systems, and improving airflow with targeted cooling and blanking panels can add significantly to the energy savings overall.
For the power & cooling energy, this calculator actually leverages the same efficiency model used in the Data Center Efficiency Calculator, which I’ve talked about in an earlier blog. Here we use that same model to establish loss parameters for six typical data center power & cooling architectures. These architectures vary in redundancy (i.e. N or 2N), and cooling approach (i.e. DX or chilled water). You would pick the one that most closely aligns to your environment and then do some “what-ifs” with changes to your power & cooling to see how the energy losses are impacted.
The space analysis within the tool is pretty straight-forward. We assume each rack is 42U tall and requires 27 sq. ft (2.5 sq.m). Pre-virtualization servers are assumed to be 2U in height, while post-virtualization servers are assumed to be 2U for consolidation ratios less than 4 and 3U for consolidation ratios greater than 4.
I hope sharing that methodology, assumptions, and data sources gives you a good sense of the tool and I encourage you to try it out if you’re undergoing a virtualization effort or thinking about it.
We also have a white paper, #118, that discusses this topic of virtualization and how optimized power & cooling can maximize the energy benefits. Take a look if you want to read more about the power and cooling challenges associated with virtualization, how to improve your PUE, impact of virtualization on redundancy requirements, and how dynamic loads can impact the risk of downtime if rack-level power and cooling health are not considered.
Remember, you can also get to our full set of Schneider Electric TradeOff Tools at http://tools.apc.com. If you have any trouble launching the tools please email me at email@example.com . As always, feedback on our tools is encouraged!