1. The bigger the data centre the more efficient it can be in utilising cooling and power infrastructure
2. The bigger the data centre the better bargaining power with the electricity companies, because they love data centre loads which are nice and stable
3. Average utilisation rates for individual servers are so low (about 10-15% – even with virtualisation), that Cloud could help to boost that by sharing loads with multiple clients

The problem with all these arguments is that they assume that everyone is at the same starting point.  We are currently undertaking a project (see previous post) to measure and report on individual service costs.   This project should enlighten us as to whether we are right to host our own services or whether we would be better off hosting them elsewhere, in the cloud.  What we won’t be doing is making any assumptions about the business case for the cloud and, in my opinion, neither should you.

Other arguments against cloud efficiency are numerous, but here is my main one:

“Cloud” data centres, however you define them, are not necessarily as efficient as a corporate data centre.  In fact, one could argue that they are much less efficient than a corporate data centre and that Facebook, Google and Microsoft, etc are examples of “Corporate” data centres and not “Cloud” data centres, since their data centres exist to promote their business models (e.g. search, email and document sharing).

In my opinion, the true cloud offerings these companies also promote should be taken separately.  Cloud computing data centres must have, by definition, capacity to grow and shrink as demand requires – this means that their utilisation will vary and their efficiency (not measurable by PUE alone) will also vary but will necessarily be poor more often than not – unless they consistently over-subscribe their capacity and prioritise certain customers over others…

Consider the analogy that is often used to describe cloud computing – an “electricity grid”.  When one plugs into a socket, more often than not electricity is supplied to power whichever device is connected.  Within limits we can power all the devices in our home from the one supply.  This is great for us but a nightmare for power station operators, who manage this daily headache consistently round the clock.  Skilled operators watch TV in order to guess when the kettle may be switched on and so try and pre-empt a massive surge in demand as households around the country reach for the coffee tin or teacup.  Do you know of any cloud providers that do this yet?

No doubt, when DCIM becomes intelligent enough and fully integrated into the data centre, with robust, mature processes to make best use of it, then the efficient management of infrastructure along with demand may become a reality.  Until then, cloud is not the “green” panacea that everyone thinks it is – far from it.

I believe that CTOs and CFOs need to demand better use of power in their corporate data centres.  Many data centre managers are waking up to this fact and in Europe, while still in its infancy, the EU Code of Conduct for Data Centres is a welcome step in the right direction to assist and encourage this activity.

Steve Bowes-Phipps, Data Centres Manager explains: ‘At the University many business systems share hardware and software to ensure we make best use of the computer resources we buy. Through the CARBS project we hope to be able to identify and monitor the individual power usage of a single service such as the Finance System.’

The year-long project that began in May 2012, will involve expertise across Information Hertfordshire and other parts of the University. Knowledge and learning gained from CARBS will be shared with the HE/FE sector.

To view the blog for this project, please visit http://blogs.herts.ac.uk/carbs/

I now have full 30 minute data that goes back to January 2011.  On a monthly basis, the PUE varies from 1.21 – 1.3, dependant on the ambient temperature.  So in the Winter we record our best PUEs and in the Summer our worst.  Winter 2011 was quite warm, so we didn’t achieve the same levels of efficiency as the year before – no month was recorded below 1.23.

• Our annualised figure for the calendar year Jan-Dec 2011 equated to 1.26
• Our saving in power costs over the projected figures pre-refurbishment was approximately 1.78mWh
• Cost of electricity was 10p, and so this represents £178,000 saving
• Multiply electricity saved by British standard for CO2 (0.527) = 937,533kgCO2
• Cost per tonne of CO2 is £12; which give us £11,250 against CRC costs
• Therefore total saving is: £189,250

Now, the total cost of the refurbishment was £430,000

Therefore the Return on Investment is approximately 28 months (2.25 yrs)!

Compare this figure with our original estimates of £40,000 savings/year and ROI of under 10 years and you can see we vastly under-estimated what could be achieved.  It is true that we also underwent an aggressive virtualisation and consolidation strategy that reduced the footprint down from 23 racks prior to refurbishment to just 13 racks in our new data centre space.  However, also note that this did not have a significantly detrimental affect on our PUE – which is odd given that our IT Load has obviously reduced considerably from expected.

We are still carrying out tweaks and changing settings in order to improve efficiency.  One such tweak is to continue to gradually increase the server inlet temperature to a dry bulb 26°C (was 18°C).  When our last old (pre 2007) server is shipped out soon, we will also switch off the humidifer.

As always, I will blog the impact of these changes as they happen.

Unfortunately, while that may be a path for the lucky, rich few organisations where money is no obstacle, most of the rest of us have a limited budget, limited space to work with and a limited timeline to deliver in.

If you’re looking to enhance the efficiency of your data centre or create capacity from ‘nothing’ in order to prevent power issues and space concerns, then here are ten tips to consider:

1. Do you have to rebuild at all?  There are many best practices that can be employed that provide fantastic returns with limited or no outlay bar some technical resource.  Take a look at the EU Code of Conduct for Data Centres.  The CoC is voluntary, but signing up to it is a commitment that encourages you to follow the best practices and report on the results in data terms back to the Commission.  Reporting data is a requirement for achieving “Participant” status, but is also a valuable exercise to help you squeeze further enhancements out of your data centre facilities.
2. Virtualise and consolidate.  Sadly, despite one of the latest high-end computer servers being a costly investment that loses value very rapidly over 3-5 years, many organisations are still buying one(or more) physical machines per business application.  This would explain why the average utilisation of servers is still languishing at 6-7% (according to Gartner).  Modern technology means there is no need to do this and virtualisation using VMWare or Solaris Containers can shrink the physical footprint of your data centre by up to 70%, while making the most of your physical tin by pushing utilisation levels up.
3. Rack positioning and containment.  This is one of those low-cost best practises that can reap big benefits.  Ensure your aisles are in a hot/cold aisle arrangement and then contain either the hot aisle or the cold aisle.  Containment can be rigid (i.e expensive) or made of plastic curtains (much cheaper) – but both equally as effective.  The hot aisle is easier as you don’t have to worry about pressure differences so much.  While you’re at it – if you use an underfloor plenum and static pressure to provide cold air to the aisles, make sure you have as many gridded tiles as you need in the cold aisle only and use brush grommets to block where cables go into the floor.  Make sure all your racks employ blanking plates where gaps between servers exist.  Check air flows with CFD (Computational Fluid Dynamics) if you have the money or using an infra-red camera with a piece of paper to check air flow direction is almost as useful in the right hands.
4. Change your cooling setup.  In older data centres, cooling can take up to 120% of the amount of electricity required to power the IT load.  Most older units of the belt-driven variety are expensive to maintain and run.  For a quick fix, most CRAC/CRAH (Computer Room Air Conditioning / Handling) units can be made more efficient by adding EC-fans.  The most efficient method and that which has demonstrated excellent ROI again and again is swapping them out for ‘free air’ units.  Often these can pay back within 18 months.
5. Another large loss of efficiency is the electrical subsystem.  Modern UPSs are highly efficient at low loads as well as when fully loaded.  If you use a 2N or 2N+1 redundant arrangement, you are unlikely to run the UPS more than half load and so this will be important.  Buy only as much capacity as you need (modular UPSs are good for this) and ensure that you run with all three phases equally balanced.
6. Zoning the data centre.  Some areas of your data centre may be high density and high heat output – put these racks together and apply a suitable cooling system directly to them.  For the rest, keep them in areas where less aggressive cooling is needed so that less power is used overall.
7. Keep UPSs and batteries out of your data centre and in their own room.  Here you can keep them at their optimal temperature range without limiting what you do in the data centre.
8. Turn up the dry bulb inlet temperature for your servers.  ASHRAE and ETSI recommend allowable ranges of temperatures for systems that exceed 30 degrees Centigrade.  Do this incrementally and keep your eye on temperature differences across the whole data centre, but it has been estimated that one degree increase in inlet temperature can contribute up to 4% of savings on your power bill.
9. Meter, Monitor, Manage and Maintain.  My four ‘M’s of data centre management.  Meter your power supplies so that you know at a minimum what total power comes into your facility, and where it gets divided up with respect to IT Load and ancillary systems such as lighting, UPS, etc.  Monitor trends of energy usage – look for anomalies and Manage those out quickly and effectively to assure continued efficiency.  Finally, Maintain your equipment to manufacturer-recommended levels.  Blocked filters impair efficiencies and can cause the lifespan of computer equipment to shorten.
10. If you do decide to rebuild.  Get some training on data centre best practices from an accredited provider.  Invite suppliers to innovate using the constraints and requirements set out in your Invitation To Tender (ITT).  Get professional electrical and mechanical advice to challenge supplier responses.  Once you’ve picked a winner – build a relationship based on partnership & trust and involve all key stakeholders in your organisation at every important turn in the project.  Create some realistic targets for the supplier to meet and hold back some payment until they meet them.

Welcome to the new-look RARE-IDC Blog at the University of Hertfordshire.

I’ve been working on quite a few initiatives recently in the data centres that I want to share with you, so keep an eye out for those over the next few weeks.  Please feel free to email me and/or leave comments on my postings to let me know if you find anything here interesting or useful – it’s always good to get feedback.

More Sustainable ICT projects are in the pipeline too, so you can look forward to hearing how we get on with those.  If all goes to plan, some really interesting work could be shaping up here that I want to ensure reaches the widest possible audience, not just in the HE/FE sector but in the ICT industry as a whole.

The Green Agenda may have slipped a bit in people’s minds due to the ongoing difficult economic situation, but a lot of sustainability efforts lead to savings in the use of power and have excellent ROIs.  So I hope you’ll also continue your efforts this year and let me know how you get on too.

Steve Bowes-Phipps MBCS CDCDP

Run by the Environmental Association for Universities and Colleges (EAUC), the Green Gown Awards (GGA) recognise exceptional environmental initiatives being undertaken by universities, colleges and the learning and skills sector across the UK. With 240 applications this year, a rise of 25% from 2010, GGA are firmly established as prestigious recognition of sustainability excellence in the further and higher education sectors.

The GGA covers 13 awards categories and the University has won the coveted Green ICT category. This recognises the growing environmental importance of ICT within the sector; it encompasses a variety of actions that help minimise energy consumption, carbon emissions, waste generation and other environmental impacts associated with ICT use.

After months of scrutiny the judges said the UH entry gave “Impressive examples of best-practice features which could easily be adapted by others.” At the Awards Ceremony a delighted Steve Bowes-Phipps, UH Data Centre Manager, was presented with the impressive GGA trophy which will be proudly displayed in the College Lane Learning Resources Centre.

Steve said: “Once again, the Data Centre Refurbishment project has been recognised as a beacon of good practice both in the industry and in the HE/FE sector.” He praised his project team colleagues from across the University saying the award was fully deserved: “This has been an important 3-year development for the University which has delivered not only a world-class green data centre but also an operationally efficient one.”

Visit the Green Gown Awards website

We’ve been making these changes blindly though as we had no meters in order to measure power usage and calculating the PUE is next to impossible as the building meters are not specific enough.

The good news is that we are finally starting to make some progress! Last Friday, we had power meters installed. We needed four: (1) Main Facility Supply, (2) PDB A, (3) PDB B and (4) Utility Board. Meter 4 captures the usage of the lighting, but also an external comms room that takes its power from our UPS – a legacy piece of infrastructure that could have been architected differently if I had been there when it was designed.

Unfortunately I don’t have a network connection Meter 1 as yet, but I do have the other three meters connected up and recording. The meters we are using are the same we’ve used elsewhere: Cube IP/400s. Does anyone know a way of capturing data from these devices automatically without manual cut & paste? If you do, please let me know. They store about 2.5 months of raw data and a year of totals for trending purposes. They can also calculate cost in monetary terms as well as carbon.

Now we can calculate our PUE and really know how our efficiency improvements are making an impact…more to follow…

The Department of Energy & Climate Change (DECC) have recently published their proposals for the simplification of the Climate Reduction Commitment (CRC) Energy Efficiency scheme.

The scheme has been criticised by many for its complexity, and as a result, the Government committed itself to simplifying the CRC and published a number of discussion papers earlier this year. DECC has now summarised the proposals for a simplified scheme, intended to be applied from Phase 2 (2013) onwards.

The most significant proposals include:

• Making the qualification process easier: Under the original scheme, qualification of organisations was based on two criteria: (i) the presence of one or more settled half hourly meters; and (ii) a total electricity of at least 6,000MWh measured to such meters. Under the simplified scheme, participants will just have to prove they use a certain amount of electricity from the qualifying meter. Whether this will differ, in practice, from the original rule is currently unknown.
• Reduce the number of fuels covered by the scheme: Currently participants are required to report on their energy supplies from a list of 29 fuels. DECC now proposes to reduce this number to four: electricity, gas, kerosene, and diesel (and the latter two, where used for heating purposes).
• Move to fix price allowance sale: The initial scheme provided for an allowance auction from Phase 2 onward. The number of allowances would have been capped following the auction, with an option to purchase additional allowances on the secondary market. Current proposals, however, would establish two sales per year, with a fix price for allowances. This removes the need for businesses to come up with auctioning strategies, although it is unsure whether there will still be room for a secondary market and how this market evolve.
• Simplifying organisational rules: Previously, participation was based on highest parent company. This caused problems to many business structures, particularly private equity and other investment funds, as it did not reflect their natural structure or processes of these organisations. Under the simplified scheme, although qualification would be maintained at highest parent company, organisations will be permitted to participate as “natural business units”. What will be considered as a natural business unit is not defined in the proposals.
• Removing overlaps between the CRC scheme and other schemes: Any organisations or sites covered by a Climate Change Agreement or the EU Emissions Trading Scheme will be automatically exempt of the CRC.

Despite numerous calls from stakeholders, DECC has decided against changes made to rules dictating the landlord and tenant relationship under the scheme.

Following this review, the Government now intends to publish draft legislative proposals in early 2012 for formal public consultation.

To view the full proposals, click here.

• Currently both universities separately and routinely incur a liability for VAT on the purchase of services, subject to a partial exemption calculation.
• If shared services are provided by a third party in the case of a separate legal entity or one university providing the services to the other) there is a potential for an additional VAT cost to be created, since the provider (if a university) would only be able to recover input tax in line with its partial exemption method and would need to seek to charge the other university on the irrecoverable VAT as part of the recharge.
• As this would be a taxable supply VAT would be charged on the total recharged amount, including the irrecoverable VAT.
• If the organisation set up to deliver the shared service employed the systems administrators who have transferred across from the institution(s) no longer requiring their services in-house, then the VAT burden has gone from zero to the double-VAT described above.

Not being an Accountant, I can’t tell you what this additional levy would mount up to, but I can assure you, that on a VAT rate of 20%, this could prove uneconomic and detrimental to the business case for a shared service initiative.  Is it any wonder that no university has done this yet!

Fortunately, a ray of light has appeared in the form of the most recent White Paper issued June 2011 by the UK Coalition Government (http://bit.ly/oBPsZE).  The White Paper has recommended reviewing the VAT burden on shared services in particular and HMRC are asking for feedback on the proposal in their review published here: http://bit.ly/paBpI8

This may be the chance to influence taxation policy that we have been waiting for and then maybe the creation of a “U-Cloud” for universities will come a step closer to reality.

Americans certainly know how to work their delegates!  Starting at 08:00hrs most days and networking drinks and nibbles going on till 22-23:00hrs most evenings, I was worn out by the amount of information thrown at me, the long days and the number of people I met from across the industry.  However, it was an elevating experience and I didn’t feel out of my depth.  I received quite a few plaudits from those who heard our story and were keen to learn more about it.

This was certainly the most prestigious award to date, but I have to say, 2 months on and I’m left with a slight feeling of “so what?”.  We haven’t gained anything more in prestige that I can ascertain, and no-one is clamouring to my door for advice, job offers, etc., so once you’ve won one award do you really need to enter any more?

Look beyond the award however at the data now being fuelled by our data centre and this tells a different story.  We are very close to meeting our PUE target, and thus our savings target of 1.22 and £39,000/year.  And I still have plenty up my sleeve to improve on the efficiency numbers over the next few months – increased dry-bulb server inlet temperatures (up to 30 degrees C), humidifier ranges 50% +/- 20% and server fan modifications.  I look forward to telling you not about the latest award we’ve won but the great news that our annualised PUE is below 1.22!  Keep watching this space.