Data centre operators have long known they must prevent contamination reaching their sensitive data processing and communications equipment. Operating experience will quickly reveal how it can cause problems ranging from disagreeable aesthetics to catastrophic equipment failure and data loss if left unchecked. Equipment suppliers are equally as aware, and increasingly refuse to accept warranty claims for equipment damaged by contamination. However, data centres that run decontamination schedules to avoid such scenarios can be pleasantly surprised to find themselves generating profits from the exercise.

This opportunity arises by addressing contamination’s impairment of ICT equipment energy efficiency. And an apparently small change on a single server can significantly impact cost and carbon footprint when scaled to large 24/7/365 data centre operations. Accordingly, decontamination specialists 8 Solutions recently conducted a Power and Energy survey on a live data centre server to investigate the actual potential for saving. The results were given to the University of Southampton’s Research Institute for Industry for further analysis, extrapolation and comment as well as authentication.

In 8 Solutions’ experience, contamination reduces energy efficiency for a number of related reasons. Dust collecting on electronics circuit boards, even if electrically and chemically inert, can cause heat build-up simply through its thermal insulating properties. Current CPU heatsink designs are also vulnerable, because they typically feature thin, closely spaced metal cooling fins.

These are susceptible to contaminant particle blockage, especially if the contaminant has significant fibre content. Fibres lodge across the fins, building a barrier that captures further particles until airflow can become entirely blocked.

This reduces the heat sink’s cooling efficiency as well as blocking any fan it carries. Restriction of intake and outlet vents can alter an equipment enclosure’s pressure curve, causing greater fan loading or reduced airflow. Energy efficiency is lost and more power consumed as the fans work harder to counteract elevated temperatures and overcome increased air flow resistance.

Therefore the survey’s basis was to measure and log the power consumption of the server ‘as found’ for a week, clean the intake fan area thoroughly, repeat the measurements for a second week and compare the results. These recordings showed an immediate small improvement after cleaning; they were then passed to Southampton University for review. The Review’s author, Professor Simon Cox, concluded that the cleaning exercise was highly effective. More in-depth conclusions appear later within this article.

The survey
The survey was conducted using a Dranetz-BMI EP1 0.1% accuracy power and harmonic analyser conforming to IEC61000-4-30, a TR2510A current clamp, a portable computer and breakout cables as required. The Table below shows the survey’s findings. The Comments that follow describe why the results had to be corrected.

Comments on corrected results
If conditions had been identical for the ‘Before Cleaning’ and ‘After Cleaning’ periods, the power and energy savings would simply be the difference between the two sets of measured results. However this was not the case, because both mains supply voltage and CPU processing load average values differed for the two periods. The University’s analysis accordingly has adjustments to allow for these differences.

In fact the mains supply voltage averaged 244.7 VRMS for the first week, and 244.6 VRMS for the second. Observation indicates that as the supply voltage drops so the power consumed increases slightly – about 0.17%, or 0.15 W for the recorded 0.1 VRMS decrease. This was compensated for by recording a corrected power consumption after cleaning of 88.59 W, and a corrected weekly energy consumption of 14.883 kWH.

Differences in the CPU loading before and after cleaning the server were also noted. Four aspects of the load were monitored throughout the study: CPU Usage, Load Average, Memory Usage, and Network Traffic. These characterize the server performing its in-service workload. By correlating and weighting each of the CPU, Load, Memory and Network Traffic data figures, the weighted aggregate load after cleaning was 8±1% higher than before cleaning. A direct relationship between processing load and power drawn cannot be inferred because this depends on the precise configuration of the rack, PSU, UPS (if any), disk and other components. Also, the efficiency of the system components will vary as a function of power.

However, taking into account relevant factors, this extra load is estimated to contribute around 1-2% to the power consumption. This leads to the final “Corrected for CPU Load and Supply Voltage” results shown in the Table above.

Other relevant factors
Southampton University’s Review commented that the methodology used in the survey was good “since measurements were actually performed on a live running server with a live running load and the actual dirt it has accumulated over time in situ”. However other data centre scenarios could yield different energy saving results.

The tests were performed within a large, highly professional co-location facility that relies on providing processing power to third party users. Accordingly, the operators expend time and effort in maintaining a clean environment, because their business so obviously depends on it. Other data centres may present a less well maintained, dirtier environment. And dirtier servers will logically benefit more from restoration to cleanliness.

Also, because of time constraints, the survey was limited to cleaning the fan intake area on the outside of the server enclosure. If the server could have been shut down for enough time to clean the fans and heatsinks within, improved airflow, lower fan loading and significantly greater power saving benefits may have been achieved.

Scaling up the benefits
The corrected survey results show a power saving of 1.65 W, and energy savings of 0.28 kWH for a single server running over one week. These figures may not seem impressive until we consider their implications for a complete data centre running over a year. If a server taking 87.14 kW can save 0.28 kWH/week, then a complete server cabinet drawing 3000 W can save (3000/87.14) x 0.28 = 9.64 kWH/week.

This equates to 9.64 x 52 = 501 kWH/year. For a data centre that has, say, 400 server cabinets, the yearly saving is 400 x 501 = 200,000 kWH. For a data centre paying 10p/kWH, this represents a yearly saving of £20,000. Reduction in CO? emissions is significant as well: According to UK Government figures, saving 1 kWH of electricity is equivalent to saving 0.43 kg CO?. The data centre would save 200,000 x 0.43 = 86,000 Kg CO? in a year.

Conclusions
The University of Southampton review of the survey results showed that, in their opinion, the cleaning procedure is highly effective. The review also concluded that both more contamination on the server and heavier process loading on the CPU could increase power demand and therefore potential for saving. A simple extrapolation from single server week to data centre year shows how significant the savings will be.

Investment in a professional data centre decontamination service will show a rapid return on investment. The CO? saving is also valuable in a climate where both legislative forces and preservation of public image put pressure onto organizations to reduce – and be seen to reduce – their carbon footprint.

ShareThis

Tags: Design & Facilities Management

Read more White Papers»

More DCS TV»

More Audio»

Read more Web Exclusives»

Read more Magazine Articles»

Read more Supplements »

Keep up to date with the latest industry products, services and technologies from the world's leading IT companies.

• Virtualization Success Depends on Network Automation by Infoblox

• The Infoblox VMware vCenter Orchestrator Plug-in by Infoblox

• Automated Storage Tiering on Infortrend’s ESVA Solution by Infortrend

• Why Cloud Backup: Top 10 Reasons by Autonomy

• How to Measure the ROI of Cloud Data Protection by Autonomy

Latest IT jobs from leading companies.

Click here for full listings»