Rack Mounted Micro-turbines for Data Centers

Microsoft researchers have proposed rack mounted fuel cells to power servers in a data centre (No More Electrical Infrastructure: Towards Fuel Cell Powered Data Centers by Ana Carolina Riekstin, Sean James, Aman Kansal, Jie Liu, and Eric Peterson, Microsoft Data Center Genome Project, November 2013). But this is not the only option for local power generation. Natural gas can also be used to generate electricity, with modified internal combustion engines. Micro-turbines are now small enough that they could be located at each row of servers, if not in each rack. This might be an interesting topic for the students doing my ICT Sustainability Course to investigate.

Syracuse University has twelve 780 kW microturbines, but these are in a centralized facility, not distributed. The units are in two groups of three each on a separate power bus. Heat from the turbines is directed to absorption chillers for cooling the servers. Recovering energy from the waste heat would be much more difficult if the generators were distributed around the racks.

Data Centre Techniques in Canberra

"Meet the DATA CENTRE Experts" is on in Canberra, 9am to 4:45pm, 7 November 2013. Topics include "Pitfalls in Data Centre Testing and Commissioning" by Paul McKenzie and John Anderson, Design Services Consultant for a Department of Defence data center.

Also:

• Residual Current Protection in the Data Centre; a necessary evil or pathway to high availability and safety? David Morley
  Data Centre Engineer, CITEC
• Do we need to protect everything equally in our DC?  Tom Townsend, Datacentre and Networks Manager, University of Canberra.
• VESNA – DCIM: What’s that all about? Brett Wickman, Director, Total Facility Services
• Why Data Centre retrofitting and refurbishing is becoming popular. Richard Wixon and Bruce Ewen.
• Encircled Flux - is it Burlesque or Broadway? Michele Hart
• NABERS for Data Centres – ‘The cat is out of the bag’ Bob Sharon
• The Data Centre Manager – why have one? Stephen Corbett, Ingenium Training Services 

Australian Energy Rating System for Data Centres

Yvette D'Ath, Parliamentary Secretary for Climate Change and Energy Efficiency, launched the NABERS for Data Centres, 8 February 2013. This is claimed to be the world’s first performance-based rating tool. The National Australian Built Environment Rating System (NABERS) is a set of tools and accreditation developed by the NSW Office of Environment and Heritage (OEH) on behalf of all Australian governments. The latest tool is for rating the energy efficiency of data centers. There is a data centre efficiency self-assessment tool for free on-line use, fact sheet, guide and Rules for collecting and using data. In addition there are data centre energy saving tips and one day Assessor training course.

The NABERS IT Equipment rating tool is the world’s first performance-based benchmarking tool to compare the energy associated with the IT equipment capacity within a data centre. Experience in developing the NABERS Energy for data centres rating tools has shown that Australian data centres have limited energy monitoring and sustainability reporting. As a result NABERS has focussed on developing simple, measurable benchmarks based on the most readily accessible data, while maintaining a robust measure of
energy performance to ensure the rating is relevant and useful for industry. An IT Equipment rating measures features that are closely related to the primary functions of a data centre (processing, storage and networking) and that all data centres provide, regardless of how they provide them. Given the current limitations in energy reporting and data collection by industry; difficulties in accessing useful work and energy consumption; and the requirement to ensure the metrics fairly represent all types of data centres functions; NABERS has developed two IT Equipment metrics:

• Processing capacity – measures the sum of the number of server cores multiplied by clock speed in gigahertz (GHz), and
• Storage capacity – measures the total unformatted storage capacity in terabytes (TB).

The NABERS performance benchmark model predicts the industry median greenhouse gas emissions for a given amount of data centre processing and storage capacity. This means that if a data centre consumes more energy than the benchmark model predicts, the site is less energy efficient than the industry median (set at 3 stars), while if it consumes less energy it is more efficient than the median. ...

From: NABERS Energy for data centres: Rating Your Data Centre Energy, National Australian Built Environment Rating System, NSW Office of Environment and Heritage, February 2013

Energy Saving for Mobile Computing Using the Cloud

Qiufen Xia will speak on "Offloading Algorithms for Mobile Computing" at the Australian National University in Canberra,11:30am, 30 November 2012.

Offloading Algorithms for Mobile Computing

Qiufen Xia

 (ANU, Research School of Computer Science)

CS HDR MONITORING

DATE: 2012-11-30
TIME: 11:30:00 - 12:00:00
LOCATION: CSIT Seminar Room, N101

ABSTRACT:
Mobile devices such as netbooks and smart phones,have made computing ubiquitous. However such battery powered devices often have limited computing power for the benefit of an extended runtime. Offloading is an effective method for extending the lifetime of mobile devices by executing some partitioned components of applications remotely(eg.,on the server in a data center or in a cloud). To achieve energy saving while satisfying the given constraints, such as execution time, many partitioning and offloading algorithms have been presented. In this presentation, I will briefly talk about some of these algorithms.

Reducing the electricity cost of cloud data centers via geographical load balancing

Zichuan Xu will speak on "Reducing the electricity cost of cloud data centers via geographical load balancing" at the Australian National University in Canberra, 11am, 30 November 2012.

Reducing the electricity cost of cloud data centers via geographical load balancing

Zichuan Xu (ANU, Research School of Computer Science)

CS HDR MONITORING

DATE: 2012-11-30
TIME: 11:00:00 - 11:30:00
LOCATION: CSIT Seminar Room, N101 ...

ABSTRACT:
Data centers, serving as infrastructures for cloud services, are growing both in number and scale. However, they usually consume an enormous amount of electric power, which leads to high operation cost of cloud service providers. Reducing the operation cost of data centers is recognized as a challenging but complex task. In this talk, we will introduce the minimum operation cost problem for fair request allocations in a distributed cloud environment where its data centers and web portals are geographically located in different regions, by incorporating the diversity of spatio-temporal electricity prices in these regions, with an objective to allocate requests to different data centers for processing while the negotiated Service Level Agreements between request users and the cloud service provider are guaranteed.

See: http://cecs.anu.edu.au/seminars/more/SID/3191

ASIO Natural Gas Generator for Backup

An interesting item which came up recently is that the Australian Security and Intelligence Organisation (ASIO) wants to store 660,000 litres of diesel fuel for the backup generators at its new Canberra Headquarters (see: "Bid to store fuel at ASIO site" by Ewa Kretowicz, Canberra Times,April 7, 2012). There has been public concern expressed about the safety of storing this much fuel near a residential area. But in addition, if they need this much fuel, how large is the carbon footprint of the building going to be?

The Canberra Hospital are proposing to install a 3 MW trigeneration plant, powered by natural gas to produce electricity, hot water and air conditioning. Such plants are much more energy efficient (and less polluting) than cola fired mains electricity. The Hospital will retain its backup diesel generators, but with the gas powered trigeneration plant providing the main power source, the electricity grid will act in effect as a backup power supply for the hospital. Perhaps ASIO should consider this approach.

A reasonable setup for ASIO would be to have a gas powered trigeneration plant large rough to provide electricity heating and cooling for most days. The diesel generators and fuel supply could be reduced in size to just provide emergency power for the ICT systems and lighting (no air-conditioning) in the unlikely event that both the gas and mains electricity supplies fail simultaneously.

ASIO could also make use of the same technology as on ACTON buses, to store natural gas under pressure. A relatively small set of tanks could be used to provide one hour of gas supply, which would be sufficient to cover most interruptions to supply.

CSIRO Requires a Modular Data Centre

CSIRO have issued a Request for Tender for a Modular Data Centre to be installed in Melbourne by 31 May 2012. The leading contender for this would be APC with their close-coupled cooling, as used at Canberra Data Centre. But any supplier will have difficulty providing a turnkey data centre within a few months.

ATM ID: CSIRORFT2011-064
Agency: CSIRO
Category: 43190000 - Communications Devices and Accessories
Close Date & Time: 27-Jan-2012 2:00 pm (ACT Local time) ...

Description:

CSIRO is seeking an innovative Data Centre solution that can readily adapt to dynamic Data Centre space, power and cooling requirements over the life of the facility. The overall solution must provide a complete turnkey solution inclusive of all site infrastructures necessary to function as a standalone Data Centre and will be assessed on its ability to be delivered as a prefabricated, preassembled and pretested suite of modules. ...

From: Request for Tender for a Modular Data Centre, CSIRO, 22-Dec-2011

Data Centre as a Service

The Australian Government Information Management Office (AGIMO), has issued a "Data Centre as a Service Discussion Paper" and invited industry comment. The Data Centre as a Service (DCaaS) facility would provide a shared data centre for smaller agencies (5o spend less than $2M a year on ICT). It is planned to have the service available in the second half of 2011, so this will have to be an existing data centre, not a new building.

As well as the idea of DCaaS, which is included in the "Australian Government Data Centre Strategy 2010‑2025", the way comment from industry is invited is of interest. As well as confidential comments by email, industry is invited to post to the AGIMO blog. There is no separate "paper" and "questionnaire" provided (no use of PDF). The paper is published as a blog entry with space to type answers to questions included.

When at the Commonwealth Schools Commission I programmed a shared mainframe computer system provided by Special Minister of State. In the main this worked well. There were some problems because SMOS was another government agency and so there was no way to legally enforce a service agreement, but if you have to take your supplier to court, you have failed.

My suggestion would be for AGIMO to tender for a basic hosting service, with add-ons as options. Later consider a cloud service. It should be noted that use of shared web based applications could greatly reduce the amount of computing equipment required and so the size of the data centre needed. My experience suggests this would reduce equipment requirements by a factor of 100, through more efficient software, more efficient data storage and sharing the hardware. In place of a data centre the size of a football field, one the size of a traditional Australian suburban quarter-acre block of land would be sufficient.

Here are my answers to the questions asked:

Question 1 of 5: Is this scope realistic and will it achieve worthwhile results? Should the scope of DCaaS include applications, and if so, which ones?

The scope is realistic. However, the short time-frame will exclude new builds.The DCaaS should exclude applications, which could be the subject of a separate process, as it raises many more issues.

Question 2 of 5: Are there any other suggestions that you think should be included?

The service should provide for at least two physical locations more than 300 km apart, to allow for backup.

Question 3 of 5: Why would one of these options be better than any other? What option would you propose?

Basic hosting would be the most suitable place to start, as it is the easiest.

Disaster Recovery Hosting & Services should be provided primarily by having at least two sites.

Full Outsourced Services, Full Managed Services should be separate services, available as an option through separate contracts not necessarily via the data centre provider.

Scalable ICT Delivery (Cloud) should be considered separately. The software issues are considerable. AGIMO should consider a suite of shared web based applications for smaller agencies.

Question 4 of 5: If presented to the market, which of the offerings listed above might your organisation bid for?

1. a) Hosting
2. b) Disaster Recovery Hosting & Services
3. c) Full Outsourced Services
4. d) Full Managed Services
5. e) Scalable ICT Delivery

AGIMO might be surprised by the large number of organisations which can provide these services. Even scalable ICT delivery is now something available using free open source software within the capabilities of the average university ICT graduate.

Question 5 of 5: What factors do you sees as leading to success or failure in the provision of DCaaS to Government. This may include contract terms, service level agreements, data security and mobility, intellectual property or data ownership.

Trying to provide more than basic hosting would make the project very complex and likely to fail. Better to start simple. It is suggested that basic hosting with at least two locations for disaster recovery be provided.

It is suggested that Full Outsourced Services and Full Managed Services be considered later as options. However, for smaller agencies it might be feasible to then step directly to a cloud based service, providing shared web based applications for agencies. AGIMO could provide some basic services for free, such as an office application suite, a social networking and email service for all public servants, and then offer additional services, such as HR, finance and recruitment management, as an optional extra paid for by corporate credit card.

Google Page Rank Sparks New Computing Paradigm

In Map-Reduce and its Children today at NICTA in Canberra, Professor Jeffrey Ullman discussed the wider uses for the approach Google uses for calculating "PageRank" for web sites:

Abstract: Since its publication by Google researchers in 2004, Map-reduce has proven to be a significant advance in programming methodology that offers resilient, easy-to-code parallel computation on a modern computing cluster or "cloud." It has led to a variety of systems that improve it in different ways. One direction is raising the level of abstraction, especially to support relational-database operations. A second direction is increasing the generality, while maintaining the programmability and resiliency in the face of partial failures. We shall review the environment of NICTA a map-reduce system, give some examples of how it works, and discuss the various extensions and the technical problems they posed.
From: Map-Reduce and its Children, NICTA, 2010

Professor Ullman started by introducing Distributed File Systems, as popularised by Google. He described how they are large (many Terabytes), divided into chunks (typically 64mbytes). These databases are usually added to (with few random updates). Chunks are replicated at compute nodes in racks connected by switches. Apart from the Google File System (GFS), HDFS and CloudStore.

On the distributed file system is built Map-Reduce, Key-Value stores and SQL. Map-reduce (such as Hadoop) is alongside the object store (key value pair).

Map reduce allows for large numbers of highly parallel processing tasks. It must allow for failure of nodes without having to restart the entire job. Previously failure could be though of as a rare event, but with tens of thousands of nodes, it is common.

Key-Value pair databases (a throwback to pre SQL relational databases?). SQL-like implementers are PIG, Hive, Sawzall and Scope.

Map-reduce has two functions: map and reduce. Large numbers of tasks are created for each to run in parallel and the results combined. Each map task operates on a distinct portion of the input file. The map tasks produce pairs of values. These are then passed to the reduce tasks, with pairs with the same "key" pass to the same reduce task, where they are combined. The result of the reduce tasks then are placed in the file store.

Clearly this design is intended for tasks such as indexing web pages for searching. It is good for matrix multiplication and relational algebra (such as a natural join). It assumes there are not many updated of data in place, just data added (that assumption works well for many web applications).

The process for dealing with a failed task is simple: restart the task. Professor Ulman commented that on real systems about 5% of failures are due to software problems (such as the version of Java not being upgraded).

Map reduce can be generalised to allow any number of tasks to be strung together. Despite the complexity, such a system can be designed to be fault tolerant: either the processing is still underway or it has finished successfully. This sounds good, but still seems to be at the experimental stage. The simpler map-reduce is easier to get to work as there are just two tasks before the results are safely stored in the file system.

It would be interesting to look at how generally applicable these applications are. Normally you would think these would work for web applications, such as indexing web pages, but not traditional databases, such as bank records. Financial records in a bank require frequent update in place of data (such as the current balance). However, banks have to keep transaction records. Normally these transaction records are seen as less important that the current updates and just something needed occasionally. But if the client has access to their bank account on-line, they are more likely to be looking at the history. This makes the transaction records important. Also "cloud" based applications may result in more aggregations of data which suit this approach.

The Professor argued that recursion is now key to web based applications, such as Page-Rank and web structure. However, this appears an area for research, not practical implementation.

Microsoft Data Centre Green Savings

The Microsoft commissioned report "Cloud Computing and Sustainability: The Environmental Benefits of Moving to the Cloud" (by Accenture and WSP, Noveber 2010) shows savings of 90% for energy use and carbon emissions where a small 100 user data centre is replaced with a cloud-based service. Savings for medium data centres (1,000 users) are less at 60 to 90% and large centres (10,000 users) 30 to 60%.

The analysis was done comparing Microsoft software, but results are likely to be similar for other software. The report did not analyse the cost implications, but as much of the reduction in energy use comes about from more efficient use of equipment there should be savings on hardware (and perhaps also software).

The report lists four factors resulting in savings:

1. Dynamic Provisioning
   
2. Multi-Tenancy
3. Server Utilization
4. Data Center Efficiency

However, these could be simplified down to just one factor: economies of scale. It is more efficient for a large number of users to hare one large system, than many small ones. However, just as the Personal Computer showed that there were advantages (and a demand for) individuals to have control over their own computer use, there is a desire for some level of autonomy. Many of the factors which make large centres more efficient could be applied to smaller installations with the application of new technology and techniques. There are also dis-economies of scale which small data centres could benefit from.

As an example, an organisation could use the same software as used by large data centres for sharing work loads. At peak times low priority applications can be slowed down. Organisations can share processing capacity without using a third party intermediary.

Small data centres can share energy sources with the organisation. For example gas powered tri-generation plants allow electricity, cooling and heating to be provided by one plant. Waste heat from a data centre can be used to warm an office building.

In addition, there are software techniques which can reduce the computing capacity needed and therefore the energy use, without the need for investment in new data centres. An example is to replace inefficient office applications with efficient web based ones. Also replacing inefficient PDF and file formats with efficient HTML equivalents would further reduce processor, networking and storage requirements. A simple approach of offering summary information can further reduce loads. This could produce a 99% reduction in energy use (and equipment requirements).

Table of Contents

Executive Summary 2
Introduction: Is the Cloud a “Greener” Computing Alternative? 3
Research Approach 4
Summary Findings 5
How Does Cloud Computing Reduce Environmental Impacts of IT? 6
Dynamic Provisioning 6
Multi-Tenancy 7
Server Utilization 7
Data Center Efficiency 8
Other Important Factors 8
Case Study – Global Consumer Goods Company 10
Conclusion & Outlook 11
Expanding the Cloud 11
Further Improvements 11
Appendix 12

From: "Cloud Computing and Sustainability: The Environmental Benefits of Moving to the Cloud" by Accenture and WSP, Microsoft, Noveber 2010

Canberra's Ecological footprint

The 2010 Walter Burley Griffin Memorial Lecture was presented this evening by Professor Brenda Vale and Doctor Robert Vale, Victoria University New Zealand, at the the newly expanded National Gallery of Australia, in Canberra. This was sponsored by the Australian Institute of Architects ACT Chapter, and University of Canberra. Professor Roger Fay (University of Tasmania) was the “respondent”.

The Vales are sustainable housing pioneers, with their 1975 book "The Autonomous House" (new edition in 2002: "The New Autonomous House: Design and Planning for Sustainability"). They developed the NABERS building rating system for the Australian Government.

In their talk the Vales looked at how Canberra would be if its citizens were to adopt a sustainable ecological footprint. This is topical as the ACT Government is holding Canberra 2030 Planning Workshops.

The lecture series commemorates the work of one of Canberra's designers, Walter Burley Griffin. Brenda Vale started by making the point that Marion Mahony Griffin also had a hand in Canberra's design (Marion is commemorated by a more recent talk series in Canberra).

After a humorous comparison of Australia and New Zealand, the point was made that both countries are large enough to support their populations in a sustainable way, with a small enough “footprint”. The fair share footprint for the world is 1/9 gha/person. Kampung Naga Indonesia was shown as an example of a stable population village.

The Vales suggested thinking of the ecological footprint as a ration book. Canberra would require a 75% reduction in resource use to meet the global ration. A 1998 study reported “Canberra's Ecological Footprint” (by: Anne Close, Barney Foran, CSIRO, 1998) . But the footprint needs crop, forest and land for energy production.

The footprint could be reduced by 50% by eating vegetarian (or less by eating Kangaroos) and 25% by eliminating proceed food. This would also make the population healthier, reducing health cost.

The Vales pointed out that smaller houses have a smaller footprint, regardless of the materials used. Also energy saving from operating the houses has less effect. They pointed out that 1940s houses were 90m2 and are now about 200m2, with fewer occupants. They showed a photo of a four generation Vietnamese family with a bicycles and small, but happy house (reminded me of Kevin McCloud's documentary "Slumming It" about Dharavi in India). At the Canberra 2030 Planning Workshop, I suggested Canberra could triple its density with smaller homes.

The Vales pointed out that the ACTION bus web site gives detailed instructions on how to catch a bus. Despite this most Canberra transport energy use is from cars. Also the low cost of air travel results in high levels of energy use. hey pointed out that Canberra was designed for trams, but they were never implemented.

Of interest was the disproportionate footprint of alcohol and tobacco. This is a current issue, which reminded me of the discussion of water use in the Murray-Darling Basin. There have been objections to using the water for environmental flow, as this would decrease food production. But much of the irrigation water is used for growing grapes for wine. There is now a wine glut with falling prices (as documented in "Structural adjustment pressures in the irrigated agriculture sector in the Murray–Darling Basin", Frontier Economics, March 2010). It seems odd to have a government policy which causes environmental damage, depresses the agricultural sector and promotes alcoholism.

Some amusement as caused by the Vale's comparison of the environmental cost of pets, with a cat equating to a Smart car and a large dog to a family size car. More seriously they suggested a city in 2050 with pedestrians and cycling, vegans, walking to work.

The Vales pointed out that “The Limits to Growth” predicted global collapse by 2050 and recent CSIRO research in "A Comparison of the Limits to Growth with Thirty Years of Reality" (Graham Turner, 2008) showed the predictions so far were accurate.

What the Vale's analysis did not take into account is different levels of wealth. It seems impractical to expect rich Australians to change their lifestyle to match that of an Indonesian villager. It would seem more feasible to adjust the footprint relative to wealth. It would be more palatable to compare an Australian to a European, or North American.

For more than 30 years the Vales have been proposing solutions based on reason. However, most of the population have not acted on it. It would seem an impossible task to convince wealthy people to give up their wealth. Rather than appealing to reason, I suggest we need to appeal to stronger motivations of fear, greed, vanity and envy. These are the motivators used for convincing people to consume more and could be used to induce them to consume differently. Wealth can be redefined in a more ecologically sustainable way and "de-materialised".

Car makers are under pressure to increase the fuel efficiency of their fleet for the US Corporate Average Fuel Economy and European CO2 emissions from light-duty vehicles legislation. Luxury car makers, including Audi and Aston Martin, have responded with compact luxury models. These are small cheap cars (Audi A1 based on the VW Polo and Cygnet based on the Toyota IQ) with cosmetic changes and accessories, to justify a higher price. These small cars are profitable for the car makers and it is better for the environment if wealthy people drive small expensive cars rather than large expensive ones.

Canberra currently has planning laws to promote large houses on small blocks of land. These laws could be changed to promote small houses on small blocks of land. If done carefully, this will still allow for profits for developers and ways for householders to compete with their neighbours for status, while consuming less.

Ecological footprint of Canberra

The 2010 Walter Burley Griffin Memorial Lecture will be presented by Professor Brenda Vale and Doctor Robert Vale, 6pm, 4 November 2010 in the at James O Fairfax Theatre, National Gallery of Australia, Canberra. Seats are limited, so RSVP the Australian Institute of Architects ACT Chapter: act@raia.com.au

The Vales are sustainable housing pioneers, with their 1975 book "The Autonomous House" (new edition in 2002: "The New Autonomous House: Design and Planning for Sustainability"). The Vales developed the NABERS building rating system for the Australian Government. Work is under-way to adopt NABERS for measuring the energy efficiency of computer data centres.

In their talk they will look at how Canberra would be if its citizens were to adopt a sustainable ecological footprint.

Last week the ACT Government held a Canberra 2030 Planning Workshop. My suggestion was to triple Canberra's population (and density) to 1 million people. With the use of sustainable technologies, the population increase could be implemented while reducing stresses on the environment and using no more land.

The 400

According to IT News "Australia needs only 400 specialists to sufficiently reduce greenhouse gas emissions from ICT ..." (in "Lecturer calls for 400 Green IT specialists", by Liz Tay, Aug 13, 2010 2:45 PM). I can't really argue with that, as apparently I said it.This was for in an interview while at Data Centre Strategics in Sydney. This is one of those cases when the journalist is better able to discern the essence of what I am talking about better than I am. What I have proposed is $1M government funding to train 400 Green ICT specialists in the top 400 government agencies and companies. The students would do one of the courses offered by ANU, ACS and Open Universities Australia, and others. The staff will then work in their organisation to reduce the greenhouse emissions of the nation by 5% to 15%.

Consolidating Data Centres Like Compacting Garbage

Greetings from Data Centre Strategics in Sydney, where I just talked on Green ICT. I was using the same notes as for my talk yesterday. But I got a little tired of hearing talks about the value of consolidating data centres and virtualising servers. So I strayed from my prepared talk and described consolidation of data centres and server vitalisation as being like using a garbage compactor.

If you have a large volume of garbage, you can use a compactor to make it smaller, but you still have the same amount of garbage, just in a smaller space. The material you have has not improved in value, it is still garbage. In a similar way server consolidation results in the same software and data running on a more compact computer system. Most of the software and data on the systems, in my experience, is garbage.

An organisation can be running dozens or hundreds of instances of web servers, mail servers, database applications and other software, when they really just need a few. Similarly most of the data in the systems is duplicated and in inefficient formats. As a result, perhaps 99% of the hardware capacity is wasted dealing with unnecessary software and data.g

The way to deal with garbage is to is sift it, retaining the small amount of valuable material and disposing of the rest. In a similar way IT people need to sift through the applications and data in their systems, to retain the small amount of value and disposing of the rest.

As an example, I pointed out that the Learning Management System used for my Green ICT course is outsourced to specialist companies. Rather than have the software run on a computer at the ANU which is running assorted other software, it is on a system dedicated to running this LMS and related software. This then allows the hardware and operating system, to be tuned to this application. Also the use of the LMS removes the need for many other specialised applications. The LMS needs a database server and a web server, but that is about all.

Similarly the course content has been carefully tuned to be efficient. I avoid using inefficient PDF documents, instead using HTML. This results in much smaller files (about one tenth the size). The LMS helps with this as it will format the web pages for printing, without the need to turn the content into PDF.

As a byproduct of this web design, the content can be displayed on smart phones and tablet computers as well as ordinary desktop computers. Along with other optimisations this results in a course which takes about 1% of the resources which would be typically required. This saves the university some money, but more importantly benefits the students, who can access the course from a remote location over a slow Internet link.

International Standards for Data Centre Metrics

Greetings from DC GreenTech in Sydney. Graeme Phillipson from Connection Research talked about recently attended an international meeting on standards for data centre metrics at the APECTEL 42 Meeting in Brunei (August 2010) and reported some results. He claimed that electricity costs in Australia would double in the next few years and that electricity was the most rapidly increasing cost for data centres. What I found interesting was that unlike most "green" speakers he did not mention climate change. This was appropriate given the audience was made up of people who run data centres which consume large amounts of electricity.

Graeme referred to the ACS ICT energy audit, which he authored. This found 2.7% of CO2e in Australia. More significant was that computers use 7.1% of the electricity in Australia. He gave a good explanation of scope 1, 2 and 3 emissions. The ACS study did not cover scope 3 emissions.

The largest ICT energy use is the cooling for data centers at 18.8%. This is followed by PCs 15.8% and Printers 15.7% and then servers 14.7%.

Graeme discussed the benefits and limitations of measures such as PUE. This measures how much of the power going into a data centre actually gets to the computer equipment. This is easy to calculate but only measures the efficiency of the data centre equipment, not what the computers actually do.

In response to limitations in PUE, the Green Grid have proposed Data Centre Energy Productivity (DCeP). This measures the ration of "useful work" from the computer to energy used. The problem is how to define what useful work is.

It seems to me that the computer industry has been struggling with computer power, in terms of the work produced, for decades. Rather than try to create a new measure, it should be feasible to create energy efficiency measures for each of the existing computer benchmarks. An overall figure could be created by a weighted average of the benchmarks.

EPA did a study which found that PUEs in real data centres follow an S curve and proposed an energy star standard in April 2010.

The EU produced a data centre code of conduct.

The Japanese Green IT Promotion Council (GiPC). The details of their DPPE will be released in a white paper next week. The Japanese approach does not include "useful work".

Graeme mentioned an "emerging Australian Standard". NABERS Energy for Data Centres work is being done. This is interesting as NABERS building standards are already in place for other types of buildings, such as offices. A standard for data centres is likely to be widely adopted as companies already understand the idea of having energy rated buildings.

Graeme is dong research for the NABERS standard and so far has data for 40 data centres. The standard was due to be out July 2010 and no is unlikely before July 2011.

On 5 April 2010 green grid announced "Global Harmonization of Data Center Energy Efficiency Metrics" with US, European and Japanese bodies has agreed to work on common international standards in a media release that US, European and Japanese bodies has agreed to work on common international standards. The aim is to come up with a global standard in 2011. In the interim PUE and DCiE are good enough to be worth going on with.

The Green Grid, the IT industry’s leading voice for advancing energy
efficiency in data centers, today announced that in an unprecedented global effort, the consortium along with the U.S. Department of Energy’s Save Now and Federal Energy Management Programs, U.S.
Environmental Protection Agency’s (EPA) ENERGY STAR Program, European Union Code of Conduct, Japan Ministry of Economy, Trade and Industry’s (METI) Green IT Initiative, and Japan’s Green IT
Promotion Council (GIPC) have collectively reached an agreement on the guiding principles of data center energy efficiency metrics. This agreement represents a significant achievement, as these
organizations continue working to ensure data center efficiency metrics, measurements and reporting conventions can be applied with clarity and consistency at the global level.

“The ultimate goal is to create a set of globally accepted metrics for data center energy efficiency. One of the first, and perhaps most important factors to successfully achieving this aim is establishing a unity of communication,” said Tom Brey, IBM representative and Secretary of The Green Grid. “The Green Grid is working with organizations around the world to develop a clear and well-defined language for the way we communicate about energy efficiency metrics, which will give us a common measuring stick for all
data centers regardless of their location. With that type of consistency, we can start driving behavioral changes in the industry.”

As the basis for these guiding principles, The Green Grid and the aforementioned organizations have designated Power Usage Effectiveness (PUE) as the industry’s preferred energy efficiency metric. The group will leverage the guiding principles to help drive a universal understanding of energy efficiency metrics and generate dialogue to improve data center efficiencies. A global task force with
representatives from each of the above mentioned organizations will continue to move this initiative forward and reconvene later this year to evaluate progress. ...

From: "Global Harmonization of Data Center Energy Efficiency Metrics", The Green Grid, 5 April 2010.

Data centre in a cave under a Cathedral

One curious location for a data centre in a cave under Uspenski Cathedral in Helsinki. The waste heat from the computers is used to warm surrounding homes. District heating is common in colder regions, but this is usually from co-generation plants used for generating electricity and heating for homes as a by-product.

Learning To Implement a Green Data Centre Strategy

I will be speaking on "Learning To Implement a Green Data Centre Strategy" at DC GreenTech in Sydney, 12 August 2010. Comments, suggestions and corrections for my presentation would be welcome.

The first globally accredited Green Information Technology course for computer professionals commenced in January 2009. The course uses the web to teach how to measure and reduce greenhouse gas emissions. Sponsored by the Australian Computer Society as part of an international program for professional education, the course now run at the masters level by the Australian National University.

Course designer Tom Worthington discusses how to build strategies for green computing, reducing the total cost of ownership with consolidated data centres and applications.

Green IT Requirements for Australian Government Data Centre Facilities

The Australian Government Information Management Office (AGIMO) has released a "Data Centre Facilities Statement of Requirements – Draft " (25 June 2010), which includes requirements for sustainable (Green) IT. The only firm requirement included is certification to the Environmental Management Systems (EMS) ISO14001 standard, or equivalent. There are no mandatory requirements for energy saving, materials recycling or reporting in the draft. This is despite requirements placed on government agencies to do so. As an example, modern data centres report the energy used by client equipment, so this can be included in mandatory government energy and greenhouse gas reporting. The draft AGIMO document includes no requirement for such reporting, so that Government CIOs would be unable to comply with reporting requirements.

A "Whole-of-Government ICT Sustainability Plan" was to be developed by the Department of the Environment, Water, Heritage and the Arts (DEWHA) in conjunction with the Department of Finance and Deregulation (Finance) by December 2009. However, the plan has not yet been released. In the interim, suitable standards and applicable laws are detailed in my book "Green Technology Strategies: Using computers and telecommunications to reduce carbon emissions". Several students undertaking my courses in Green IT have written green ICT reports for their agencies as course assignments. The courses are available through the Australian Computer Society, the Australian national University and Open Universities Australia. Currently I have students studying Green ICT, including several preparing reports for major government agencies The next course starts at ANU in a few weeks time and enrolments are now open. The course will be updated when the Whole-of-Government ICT Sustainability Plan is available. Also sustainable IT is one of the topics for the World Computer Congress 2010, to be held in Brisbane in September. Several of the architects of Australian green IT plans and training will be speaking there (I will be talking about green IT training for professionals).

A Request for Tender (RFT) is planned to be issued for a panel of suppliers for the bulk of data centre facilities to the Australian Government agencies. However, as the ICT sustainability plan is now six months late, it is not clear when the RFT may be issued.

There are several interesting aspects to the draft document: apart from the large scale of the services required, the way the draft was released is of interest.

The draft requirements was announced online by Kayelle Wiltshire, Assistant Secretary of the Central Facilities Branch,AGIMO, in what is essentially a first person blog, without the usual third person, hard to understand official bureaucratic language. Comments are invited to the blog.

AGIMO are acknowledging and referencing a previous Australian Information Industry Association (AIIA) "Whole-of-Government Data Centre Strategy: Industry Best Practice Principles" (AIIA, 29 May 2010). It is unusual for an agency to acknowledge the input of an industry body so directly.

What also makes the AGIMO data centre initiative different is the use of HTML. Usually government reports are provided in large, hard to download, hard to read PDF documents. In the case AGIMO have provided web page versions of not only its own documents, but the AIIA report as well (the documents are also offered in PDF and RTF formats. One improvement AGIMO could make would be to list the HTML version first, rather than last. May people are likely to click on the first link they see and end up with a PDF file download they did not really want.

Green IT for Australian Government Data Centres

The AIIA report cited by AGIO includes a section on "Green IT':

Government should be clear on its commitment to ‘Green’ technologies and encourage innovation and commitment from industry in this area. Climate factors of a location that enable efficiency should be included in the consideration.

The AIIA report cites:

1. "Data Centre Certification Ramps Up", Bianca Wirth, GreenITStrategy.com, 6 April 2010
2. Data Centre Design Guide, Presentation by Jon Haas and Tony Pierce, The Green Grid, 4 February 2009

The draft statement of requirements includes a section on "Sustainability":

8.28 The Australian Government encourages “Green IT” and other environmental initiatives where possible, such as the reduction of greenhouse gas emissions, use of fossil fuel energy, use of water and e-waste.

8.29 The Tenderer should provide details of any Green Power (from renewable sources) used in data centre facilities.

8.30 The Tenderer should be certified to Environmental Management Systems (EMS) ISO14001 or to an equivalent environmental management standard, such as the Eco-Management and Audit Scheme (EMAS). If the Tenderer is not certified, the Tenderer should commit to be certified within 12 months of joining the panel of data centre Providers.

8.31 The Tenderer should provide the current greenhouse gas emission baseline for data centre facilities and plans, if any, for upgrades to infrastructure to reduce the greenhouse gas footprint.

8.32 It is desirable that the data centre operation minimises the use of water and power. The Tenderer should detail how this can be achieved.

Green issues are also mentioned in the section on Un-interruptable Power Supplies (UPS):

8.16 The Tenderer should provide evidence of green initiatives in UPS design that demonstrates the use of non-chemical based UPS systems and the reduction in the use of harmful chemicals in the UPS.

Green data centres in Canberra

Greg Boorer, Managing Director of Canberra Data Centres (CDC) will be speaking on the “Pod Architecture” for greener data centre facilities, at ACS Young IT in the Pub, 25 June 2010. I had a tour of Canberra Data Centres last year.

Shred Disk Drives

In his talk "Overlooking the Obvious - Security Blunders - Costs, Cutbacks & Catastrophes" to the Australian Computer Society in Canberra this evening, David Cook, Manager, SECAU Security Research Centre, recommended organisations shred disk drives before disposal. David cited research by Edith Cowan University and University of Glamorgan, indicating that one in six of the disk drives disposed of still contain easily recovered data. He argued that using a disk wiping program was not sufficient for sensitive corporate and government data: the hard disk drive should be reduced to dust to ensure the data is destroyed.

Yesterday, the Australian Government announced funding for the initial stages to implement a whole‑of‑government data centre strategy for the procurement of data centre facilities. Given the high risk of the loss of sensitive government data and the low cost of disk drives, perhaps the government should institute a policy that no disk drives will leave the data centres. Shredders could be installed at the data centres and security staff trained to destroy all surplus disk drives and other data devices.

David also mentioned very recent work (not yet ready for publication) which indicated that one in three mobile phones had easily read data on it. He pointed to Blackberry smart phones, which have a reputation for security, but where many users fail to turn on the security lock. Blackberries are in common use by government. Perhaps these should be configured so the security lock cannot be switched off.

David also expressed concern about some universities outsourcing their email to a "cloud" service, with it not clear where the data was held. He argued that new laws were needed to cover this and other privacy issues. However, existing laws would cover many privacy issues. As an example, universities are required to comply with codes, such as "The National Code of Practice for Registration Authorities and Providers of Education and Training to Overseas Students". Those universities which do not protect their students privacy can be delisted. Similarly, banking, insurance and medical data is protected under Australian law. In the case of state and federal government agencies, criminal penalties apply for the misuse of data.