Ok, so in the previous post I mused on the Data Centre (DC) of the Future (DCF) and confirmed that in my opinion, we already had the Strategic, Cultural and Environmental direction already laid out for us to follow in the EU Code of Conduct for Data Centres (Energy Efficiency) best practice guidelines. I hadn't considered exactly how many though! Thus I've had to split them into sections, so this series has just got considerably longer, it'll probably be published in 7 chapters or even put out in a ebook (I think I might charge for it though, comments welcomed on how much and what readers would really like to see in it!)
But, and its a big but, there are other Standards, and potential regulations that need to be considered. Over the past year I have been working under contract with Green IT Amsterdam http://www.greenitamsterdam.nl on the CATALYST Project www.project-catalyst.eu and we have created the Green Data Centre Roadmap which can be found here https://roadmap.project-catalyst.eu/ This map, together with the Green DC Assessment Tool, Handbook and Directory provides most of the information a data centre design, owner, operator and customers can use to be as energy efficiency and sustainable as possible.
Most of the Global Data Centre standards and some recommended guidelines are included on the map, together with some information in the Handbook, however this is a moot point as the ISO have published the TS22237 series, which are the EN50600 series, parts 1, 2-1 - 2-5 and 3-1, 7 Standards altogether and as its an ISO will be applicable globally in the future. Currently its a Technical Specification, and this means in effect that it is a 'consultative document' and a prelude to the creation of a Full ISO Standard.
We also have to consider the EU Green Deal, back in February/March, the EU commented that 'Data Centres, can, and should be carbon neutral by 2030' and there is an underlying threat of regulation, we must accept that the EU will be serious in this, and it is possible that they will use the EU Code of Conduct for Data Centres (Energy Efficiency) as their primary tool to ensure that Data Centres meet the carbon neutral goal.
So, what does that mean? Well, they may issue a Directive, requiring all DCs over a certain size to be particpants to the EUCOC, (which National Governments will need to implement via primary legislation by a certain date, and given the 2030 deadline, one would hope that this would be sooner rather than later), in order for data centres to begin their journey to carbon neutrality. This of course, then raises the issue of compliance, currently, DCs registered as participants are required to submit annual updates on energy consumption and progress on their action plans annually, would this be sufficient? There are approximately just under 400 DCs registered as EUCOC participants, would a mandatory requirement hugely inflate the register?, well yes, other studies have summised that there are 100,000 plus DCs in the EU, although some may not be of sufficient size to warrant compliance. Does this mean that a data centre compliance office would need to be created?, If so, what powers would they have and how would it be funded? Would it be at EU level or National level? All questions the EC will have to address and pretty soon.
Of course, the UK would not be required to participate, but does that mean that UK data centres escape?, I suspect that the UK Government would closely watch activities on the other side of the Channel and probably implement something similar.
Of course, this is just me 'musing' on what could happen, it may never happen, but operators need to be prepared!
Of course, there are other methods of steering data centres into being energy efficient and sustainable, this could come from other 'energy' legislation, such as the EU Energy Efficiency Directives, which require 4 yearly audits, unless you are exempted via an ISO50001 Energy Management Systems certification. This is a source of some angst within the DC sector, as there is a requirement to 'continiously improve' energy performance, this is at odds with standard data centre practice, where there is an energy consumption curve as you fill out the facility which only ever goes in one direction, up!
Other steers, can be found in national, regional or local planning requirements, for instance, the Netherlands has a regional spatial strategy, data centres can only be built in 6 specific areas and there are local planning guidance that applies as well, this guidance includes, energy efficiency, water use, connection to district heating networks and nature aspects (bird and bat boxes!) and in some areas, additional requirements can be required, such as the Schiphol Airport Development Company's Green Data Center Campus Sustainability Framework for Data Centers. [Note, that further information on all these can be found on the CATALYST Green Data Centre Roadmap mentioned earlier!]
So, what does this all mean for the Data Centre of the Future?
Well, one thing that springs to mind is that current cookie cutter designs really aren't going to cut it, unless they have covered all the potential issues that may arise and from personal exprience we're still along way from that.
I noted in a linked post recently that colocation sites are very much customer driven, insofar that they have to provide the infrastructure that their customers require and not some future minded Green IT and Data Centre energy consultants wish list (ahem) but I disagree.
Back in mid August, I was a panelist on the Confederation of Indian Industries, Data Centre Energy Efficiency Workshop and earlier in the session, in the introduction, one of the presenters said that 'Everybody needs to get Green' IT Manufacturers, M&E suppliers, DC owners and operators and customers have to work together to implement energy efficient, sustainable and carbon neutral IT, and this starts with the ICT manufacturers, quite simply they have to step up and redesign their equipment to use less energy and natural resources, but still provide the services that people buy them for, in addition, they need to start providing equipment in two flavours, one for air cooled systems and one for liquid cooled systems.
This is a fundamental design issue, chips are analyzed using CFD to determine air flow, as almost everything is designed to be cooled by air, then the servers themselves are also analyzed using CFD to determine airflow across the machine itself, however, most data centres are not CFD analyzed at design stage, simply because the designers have zero visibilty of the ICT equipment that will be installed or their airflow characteristics. What does this mean? Well, DC designers have to make assumptions on air flow, and rely on a basic airflow cycle (That is as follows, air leaves CRAC/H and is pushed underfloor, air escapes floor void at grills (cold aisle) in front of racks, air rises and is drawn into the racks by the fans in each server, air passes over the heat producing components (chips and power supply) takes the heat and it is exhausted to the rear of the server (hot aisle), hot air rises and is drawn back into the top of the CRAC/H, passes over the coil, transfers its heat and the cycle starts again. The heat is then taken outside and vented to the atmosphere (yes I know there are other methods!) This method is not sophisticated at all, it is simply the movement of air, this is one of the reasons why some legacy and even newer facilities develop 'hot spots' and can use more energy.
Another issue arises and it is that air is not a great carrier of heat, and if we want to use data centre waste heat in any great amount, another solution has to be found.
That solution unsurprisingly is liquids, now DCs already in some cases, transfer their heat to liquid solutions, indeed a CRAC can transfer the return heat to the coil and the coil is liquid, this is in essence a chilled water loop, but there is a disconnect, hot air is at its hottest as it leaves the server (some 50degrees C) but it will lose heat through time and distance, before it arrives at the CRAC coil, some 40% in fact! as Return temps at CRACs are around 30degreesC
If you want to learn more about 'Air v Liquid' then catch up on our previous series on this topic which can be found on the following links Air v Liquid Part 1, Air v Liquid Part 2, Air v Liquid Part 3 and the Air v Liquid Part 4.
So, now I'm going to cover the first section of the EUCOC Data Centre Utilisation, Management & Planning and the best practices that point towards that 'strategic and cultural change' I mentioned earlier.
Before I cover this, I'd just like to make a serious point, currently ALL IT services are, in Data Centre Tier and EN 50600 Class levels, the same, meaning that a P1 application has the same data centre resilience as a P4 service, this is because the DC and its design is to the UTI Tier Topology or EN 50600 classes.
EUCOC 3.1.1 Group Involvement - So this best practice appears to be merely an 'approval board' and perhaps the creators of a 'standard hardware list' however, this statement 'Require the approval of this group for any significant decision to ensure that the impacts of the decision have been properly understood and an effective solution reached' really points towards strategic change for our DCF, this group, in my opinion needs to include senior management and not just the representatives from IT (I might push for this at our next meeting!) to provide some direction for our DCF which of course has to include the 'on prem v cloud' argument, a redefinition of what are the 'critical services' and the levels of resilience that should be required. And one final point, should this board be facilitating the move to an energy efficient and sustainable DC facility? In my opinion, YES it should!
EUCOC 3.2.3 Service Charging Models - Currently, most enterprises, and to a certain extent colocation facilities, have a very simple pricing model, for Enterprises, it really is a 'one size fits all' all DC services and the cost of those services is usually covered at a corporate level meaning that the cost of providing all the IT is divided by the amount of users and applied to each business unit, so, for instance, the SALES department with 100 users costs X x 100, whilst the FINANCE department with 5 users costs X x 5, However, the Finance department has more equipment in use and costs more in terms of energy, there is a fundamental disconnect. This is then compounded as each business unit has its own budgets, which can be used for Shadow IT (IT Services provided by others with no direct contract or involvement by the Corporate IT department, examples include the use of 3rd party cloud services) or the procurement of equipment that is not on the standard hardware list (see 3.1.1) al this leads to reduced visibility within the corporate IT department and the DC element, so control is lost.
Put simply the DC Manager is reactive rather than proactive, and costs rise as a result. Another example is the use of colocation services, this is usually priced on a 'per rack basis' together with a energy per kW so, a single rack with one server will cost X (Rack Monthly Charge) + Y (Cost of Energy) + Z (Network Charge, which could be both the cost of the circuit, plus inbound/outbound data traffic charged by the MB/GB/TB etc). Now, its probably simpler to adopt this approach as it is in effect a 'shared service' both the sales and finance users are using the rack, but in effect the sales department is subsidizing the finance department in terms of the overall cost.
All together this reduces visibility of the actual cost of providing IT to each department means that the job of making the facility more efficient is more difficult.
The resolution is to calculate the actual cost of providing those DC services to each department and then charge them, that way the user is intimately involved and you will see change, cultural change, no department wants to be seen as the 'energy guzzling, unsustainable' department within the business!
EUCOC 3.2.4 Life Cycle Assessment (Optional) - the Best practice suggests the introduction of a plan for Life Cycle Assessment (LCA) in accordance with emerging guidelines (In my opinion this probably needs to be 'beefed up' at our next EUCOC meeting as the guidelines are no longer emerging!) and this in effect requires the calculation of both embodied energy (of the device or component, server, network switch, CRAC, Chiller etc) and its 'in use' phase.
There is plenty of material available to complete this best practice and it is one of our core skills (contact us for further information!)
But, once again this is strategic and cultural change element, if you include the concept of 'carbon budgets' for IT services, then you need to find out what embodied energy there is throughout your IT estate, this can only be provided by the suppliers of those services, to date, asking for a LCA sheet from a manufacturer has proved to be problematic, but the more people that ask, then the more pressure will be put on manufactureres to start to calculate them. If you insist that the overall carbon/energy use is calculated and then applied to a service, then your 'designer/architect' has to comply!
EUCOC 3.2.5 Environmental Management - This should be a given, public sector authorities usually have this as a pre-requisite in any public facing tender, hence the reason why colocation operators seeking to provide services to the public sector usually have the accreditation (usually ISO14001). However, for Enterprises, DC's are usually missing from the scope, its time to get them included, especially as a DC will probably account for 25% plus of energy use within an organisation.
EUCOC 3.2.6 Energy Management - This can be a problem, why? Well, the focus for ISO50001 Energy Management Systems is on the overall reduction of energy for an organisation, this can be useful for Enterprises, but not so much for Colos where their electricity demand curve should, if they are attracting and retaining customers only be going one way, Up! That said the requirement for lots of energy data and that decisions should be made with energy efficiency in mind also makes this a strategic and cultural change driver!
EUCOC 3.2.8 Sustainable Energy Use - This best practice requires that you ' Consider the proportion of energy used by the data centre that comes from renewable / sustainable sources'. And that 'Recording and reporting on the proportion of sustainable / renewable energy used against the overall energy consumption is expected to become an expected monitoring and reporting requirement in time'. There are a whole bunch of DC KPIs included in the ISO 30134 series and the EN 50600 4-X series that can be used, this is one for the DC Committee to consider as there will be a requirements to be involved in the external POWER PURCHASE AGREEMENTS (PPA's) and the procurement, installation, commissioning and USE or advanced energy monitoring solutions, such as Data Centre infrastructure Monitoring (DCiM). Again, this is a strategic and cultural change, you may wonder why I capitalised the 'USE' of DCiM in the sentence above, its because far too often, the DCiM is not used to its full potential, and considering the time, effort and cost of the solution, its silly not to use it effectively, this is primarily a cultural thing!
EUCOC 3.2.11 Alternative power generation technologies (Optional) - This best practice requires that ' Consideration should be given to energy supplies from alternative and sustainable energy sources including fuel cells, wind power, photo-voltaic and shared local-generation using biomass / bio-fuels etc., which might provide a lower carbon footprint, economically attractive and reliable alternative to utility electricity supplies. Now, if this isn't a strategic and cultural change driver, I dont know what is!
EUCOC 3.2.15 Training & Development - Now, in the UK, Training is sometimes given as a reward or as a tester for the organisation, but its usually the most senior tech that goes on the course, he returns, then tells his manager, that he didn't learn anything (this is probably untrue, but its a common occurence!) so the manager doesn't send the very people that DO need the training. This certainly needs to change and as the best practice states ' An overall training methodology should be defined and delivered to all relevant personnel in order to improve individual personal development and develop suitably qualified and experienced data centre personnel.'
As an aside, Carbon3IT have a range of training courses, and we are actively looking into remote options, see our training page here. We also provide ad-hoc courses designed specifically around your environment.
Of course, this is another strategic, as in you will need your tech staff to be fully conversant with some of the newer technologies and concepts being developed and these are usually delivered by specialist training courses, incidentially the CATALYST Roadmap has a training line with some of the larger training companies listed
EUCOC 3.3.1 Build resilience to business requirements - So, another strategic change element, in effect, everytime you deploy a service, question the architecture, question the 24/7/365 aspects, does the service really need to be available for 24hours a day, 7 days a week for 365/6 days a year, could it be changed to business hours, say 7am - 7pm, 5 days a week and not available during seasonal holidays? Or, subject to use and traffic factors, spinning up when needed? In essence, this best practice is giving you the ability to question the entire rationale behind your infrastructure and giving you a great business case for a systems audit (other best practices indicate the need for regular audits!)
EUCOC 3.3.2 Consider multiple levels of resilience - If you recall BP 3.2.3 you will recollect that the Service Charging Models give you the opportunity to apply the true costs of services to individual departments and users, this will make them 'think, and 3.3.1 above gives you the option to question them on their stated requirements for their services. This, of course, then gives you the opportunity to real tailor your DC services to the actual applications in use within the organisation and 'their' criticality to the business.
An example, I was doing some BC/DR planning for a large retail organisation and asked what 'they, as in the IT management team' considered as their most critical application, the answer C-Suite email, No, no, no, I said, your 'most critical applications are your EPOS systems, so customers can buy the items you have on the shelves, and your logistics system, so that items can be replaced on the shelves quickly, these are your mission critical applications as they bring in revenue, the C-Suite, if their email goes down can pick up the telephone or get off their chairs (I used a slang term for a 'behind') and walk around to whoever they need to send an email to, they could probably do with the exercise and its better sometimes to discuss things face to face. Anyway, my point here is that every service needs to be reviewed and a grade applied so to its (the service) criticality to the business, this will obviously change depending on what the organisations 'core' business is.
EUCOC 3.3.3 Lean provisioning of power and cooling for a maximum of 18 months of data floor capacity - This best practice is interesting, because, many DCs have some strange design principles applied, the structure will probably be designed to last 75 years, the M&E infrastructure 25 years, the IT systems anywhere between 5-7 years depending on the organisation and the technology roadmap, so within the intended lifetime of the building, the M&E will be replaced 3 or four times, and the IT at least 13 times, however history tells us that IT is getting smaller and more dense, so why spend money on an external shell and core that fairly rapidly will be obselete? This is certainly a strategic element, choose the best option for your organisation, but look to the future, there are plenty of examples of a 'state of the art' DC being considered 'legacy' only a few years later.
EUCOC 3.3.4 Design infrastructure to maximise part load efficiency - This best practice is at odds with the previous one, and thats because we know that you won't take any notice, and still design DCs the way you have always done, so this one says, if you are going to ignore us and build as you have always done, then please, please, please, make the design have the ability to shut down for part load, there really is no point building a huge 10MW facility for an initial 200kW load, without giving yourself the ability to shut down bits of it, when its not in use.
EUCOC 3.3.5 Design appropriate levels of resilience (Opt) - This best practice is a continuation of 3.3.2, if you have considered 'multiple levels of resilience' then the next step is to 'design appropriate levels of resilience, note that this best practice links in with the various audit and assessment best practices and in line with 3.2.3 Service Charging Models.
So, thats Section 3 complete, remember that we've only been looking at the STRATEGIC & CULTURAL change elements and not the technical elements, although in some cases they do have some bearing.
Please feel free to post any comments on my thinking, I'll post links to the post on linked in as well, and you can comment there as well.