Message from the Industry Chair – Anne Holler, VMware

I appreciate the chance to serve as Industry Chair for the IEEE CS STC on Sustainable Computing, following Canturk Isci's strong contributions in this area. The work in this vibrant growing community is inspiring! And thanks to Diwakar for the invitation to submit an editorial piece.

I think a key opportunity for sustainable computing is to do the hard engineering to make it easy  to be green, ideally, easier than less green alternatives. As an analogy, driving a hybrid car is  typically considered greener than driving a comparable gasoline-only car. But a really interesting thing is that, in my experience, driving a hybrid car is actually easier than driving a gasoline-only car, given the significantly fewer trips to the gas station. So, in this case, it's actually easier being greener, but it took significant innovation in automotive engineering to make this so.  Working in the area of datacenter resource management at VMware, I see technology examples where it is easy to be green and examples where there is opportunity for improvement.

The transformation of many mainstream enterprise datacenters from running workloads directly on lightly-utilized dedicated hosts to consolidating those workloads in virtual machines (VMs) running on more highly-utilized servers was greener, and many power companies gave rebates to reward the savings. But a really interesting thing is that the transformation made datacenter management easier. Deploying
new server workloads was quicker and easier in VMs than on physical servers. Live migration allowed maintenance of servers and datastores to be handled without downtime and enabled the automatic scheduling of resources to VMs, further increasing consolidation while supporting service-level agreements, fairness, normalized headroom for demand spikes, and reservation of spare resources for failover.

However, at this point, some datacenters are not continuing to increase VM consolidation per host, due to concerns including the impact of resource overcommitment on application performance. Some are not adopting power proportionality technologies, due to the risk of added latency in response to demand spikes. Some are underutilizing power capacity, due to conservative assumptions about
server peak power draw. For these cases, I would argue that the academic and industry solutions  (certainly including those from VMware!) have not fully reached the bar of easy being green; the solutions require manual tuning or do not interoperate seamlessly with other key datacenter software.

To push datacenter resource efficiency forward from this point, I believe the following are crucial:

• Machine learning
•  Integrated use of analytics
•  Deeper interoperability in resource management

At VMware, we are working on using machine learning to manage resources to maintain application performance targets, which avoids manual tuning of low-level resource controls. We are exploring the use of analytics to automatically identify demand patterns to support proactive preparation for expected demand spikes. And we are continuing to deepen resource management interoperation, including investigating rack power budget management across a set of hosts running VM workloads in accordance with cloud resource management system.

I'd be interested to hear "easy to be green" examples from across the Sustainable Computing community, as well as particular examples where we've made strides but haven't yet completely reached the bar.

-Cheers, Anne

Community Highlight

Interview with Dr. Sven Beiker by Lei Rao, General Motors Research Lab, USA

Dr. Sven A. Beiker is the Executive Director of CARS, the Center for Automotive Research at Stanford. CARS is an interdisciplinary partnership between academia and industry to address the challenges of personal mobility in the 21st century. Since 2008, Dr. Beiker has been taking care of the program’s strategic planning, resources management, and project incubation. He was instrumental in launching research programs at Stanford University in "Legal Aspects of Autonomous Driving", "Wireless Power Transfer to a Moving Vehicle", and “Vehicle Communication via Cellular Networks” as well as "The Revs Program at Stanford" and the "Volkswagen Automotive Innovation Lab".
 
Personal webpage: https://profiles.stanford.edu/sven-beiker
 

Prelude:
 
The automotive industry has been leading the technology and innovation to reduce its environmental impact as well as minimizing the costs and resource usages. The major topics of sustainability in traditional automotive industry include but do not limit to improving the efficiencies of vehicles, vehicle recycling and real-time driving safety and comfort enhancements. Nowadays safety and efficiency in road traffic have been considered as top priorities. Automated driving technologies have the potential to make mobility optimally safe and sustainable in the future. We spoke to a distinguished researcher, Dr. Sven A. Beiker, in the automated driving research area to find out more about the field.
 
Favorite memory as a researcher/Executive Director of CARS
 
The most enjoyable part is to have the great opportunity to discuss with a diverse spectrum of people, including students, faculties, industry experts and top-level executives. It is a great experience to learn from them and discuss with them on what the future of automotive industry might be. This broad variety of people sometimes can bring innovative thoughts to the automated vehicle research.
 
 
Opinions and comments towards green vehicular systems; impact of automated driving towards a more sustainable vehicular systems
 
It is important to understand that there are different perspectives of sustainability, when we talk about vehicular systems. In my opinion, the most important two perspectives in this regard are safety and environment. Automated driving technologies and systems definitely help to improve the safety and driving experiences for drivers. There exist some concerns about the environmental perspective. If the driving can get much easier, people may end up with driving a lot more when they do not need to think much about the driving. It is not very clear at the current stage whether automated driving will help reduce or increase traffic overall. An existing challenge for us is to help propose solutions to achieve a balance between driving efforts and driving distances.
 
 
Most promising applications/research in autonomous driving for sustainability
 
As mentioned above, safety and environment are the most important two perspectives here. Collision avoidance technologies and distracted driving solutions are most important for safety. Traffic control and optimization technologies can be used to improve the environmental aspect. Good examples include methods to respond to traffic dynamics and enable closer follow distances among vehicles*.
 
Brief takeaways for students and young researchers
 
It is important to apply interdisciplinary thinking in doing automated driving research. Good researchers cannot only focus on traditional handbooks and theories of automotive research. Keep eyes more open and think broadly.
 
* Note: computer-assisted and networked driving systems can react to traffic dynamics and potentially dangerous situations much faster and more precisely than a human being, thus enabling closer following distances among vehicles to increase the road efficiency and at the same providing safety guarantees.

Privacy options for the Automatic Metering Infrastructure of Smart Grids

by Cristina Rottondi, Politecnico di Milano

The amount of user data collected by the Smart Grid is expected to dramatically increase with respect to the current electrical power grid: this creates great concerns regarding the privacy of the customers. The new Smart Meters will provide to the Smart Grid not only nearly real-time information about the energy consumption, but also a great amount of user-related data which will be used by the utilities themselves (e.g. for billing purposes), the grid managers (e.g. for electrical power state estimation) or third parties (e.g. to provide value-added services, such as home energy consumption management).

Since the Smart Meter’s location in the user’s household makes it an easy target for tampering attacks, a first security challenge is to provide them with security mechanisms capable of making the device tamper proof, in order to ensure that the gathered readings are not altered, leading to incorrect billing or to wrong estimations of the power usage. The goal of this kind of attack can range from monetary gains (it has been estimated that 6 billion of dollars worth of power is being stolen from the U.S. electrical power system [1]) to Denial of Service (DoS) or other terroristic attacks. The most widely adopted countermeasure is to include a Hardware Security Module (HSM) in the metering device, which can provide both physical and logical protection to cryptographic and sensitive data (e.g. cryptographic keys) and even execute specific algorithms which require a secure and controlled environment (e.g. a billing protocol).

In addition to integrity, data confidentiality must also be ensured. Since Smart Meters are connected to the Smart Grid communication network to send their readings to the power suppliers, external subjects might access these data and infer private information about the users. To tackle such privacy issues, Cavoukian et al. [2] propose the “Privacy by design” paradigm and suggest several guidelines to utilities and operators dealing with smart metering data. The underlying basic criterion is that any entity providing personal information related to energy consumption data to a third party must obtain in advance the express consent of the individuals whose meter readings are collected.

However, designing a privacy-friendly infrastructure for metering data collection is still an open issue, due to the criticalities of the Smart Grid environment, including scalability, reliability and fault resiliency.
Several approaches have been proposed by the scientific community (the reader is referred to [3] for a comprehensive overview). The most widely investigated are:

• entrusting the Smart Meter with performing calculations and providing the backend system with the results, using cryptographic commitments and Zero Knowledge Proofs to verify the results in order to prevent the meter from cheating. Such approach often requires the usage of the previously mentioned HSM.

• using MultiParty Computation (MPC) to perform the collaborative computation of an aggregation function, generally a sum, over the meter readings without directly exposing individual data. MPC is a field of cryptography which designs algorithms requiring the participation of n multiple entities, each one performing cryptographic operations over its own input data d1, …, dn to calculate a generic function f(d1, …, dn) without leaking any other information (e.g. the individual inputs themselves or the result of any intermediate computation). Note that the MPC approach can be distributed over all the users or can exploit one or more servers. However, the privacy guarantees of this approach are unclear, since the privacy of a single user depends on the behaviour of the other users. For example, let us assume that meter readings generated by all the households in a block of flats are aggregated before being communicated to the grid manager for monitoring purposes. If the number of households is sufficiently large, the fluctuations in the individual consumption patterns have a marginal impact on the aggregate. But what happens if during the day most of the households are unoccupied? Clearly, their energy usage will drastically decrease, thus exposing the individual patterns of the remaining apartments and diminishing their privacy level;

• *hiding the identity of the subjects by using pseudonyms, so that data can be delivered to the utility or third party without revealing the identity of the user who generated them. The drawback of such anonymization techniques is that they cannot provide accountability, which makes them useless for several kinds of applications, including billing;

• performing noise addition on the collected data, in order to increase the failure rate of de-anonimization or non-intrusive load monitoring techniques. However, such an approach makes the metering results less useful for the legitimate recipient. Dimensioning the noise to be added to the aggregate is a crucial issue to make privacy-friendly approaches feasible for deployment in a large-scale grid.

It follows that none of the above mentioned techniques can be considered as a panacea for the privacy issues of the Smart Grid scenario: only combining such methods in a comprehensive framework can provide a sufficient level of privacy protection for the users without compromising the effectiveness of the management of the electricity grid of the future.

 
[1] P. McDaniel and S. McLaughlin. Security and privacy challenges in the smart grid. Security
Privacy, IEEE, 7(3):75 –77, may-june 2009.
 
[2] Ann Cavoukian, Jules Polonetsky, and Christopher Wolf. Smartprivacy for the smart grid:
embedding privacy into the design of electricity conservation. Identity in the Information
Society, 3:275–294, 2010.
 
[3] M. Jawurek, F. Kerschbaum, and G. Danezis. Sok: Privacy technologies for smart grids – a
survey of options., 2012.

Sustainable food through information technology

by David Aikema, University of Calgary, Canada

In 2010, HP Labs' "Design of Farm Waste-Driven Supply Side Infrastructure for Data Centers" [1] drew the attention of the sustainable computing community to the potential of farms to drive computing in a more sustainable direction. The paper studied how a bio-digester fed with farm waste could be used to power a data centre. This article highlights another facet of the link between information technology and farming: the significant potential for computing to improve the sustainability of the agricultural and broader food-related sectors of the economy.

As addicted as many of us are to computers and computing services in their various shapes and forms, food is even more difficult to abandon. Unfortunately, the agricultural sector is also one of the largest sources of emissions, estimated to account for 10-12% of global anthropogenic emissions in a recent IPCC report [2]. With the population anticipated to continue increasing for the next few decades and individuals on average consuming more animal products as their wealth increases, agriculture faces significant challenges in the coming years. Expert testimony to the UN predicts that food production will need to double by 2050 [3].


The use of information technology is not alien to the agricultural sector. There's been significant technological development in the agricultural sector over the past decade though it has happened largely disconnected from the sustainable computing community. In spite of stereotypes of farmers being technologically backwards, technologies such as auto steering of farm vehicles have become relatively common [4] far in advance of self-driving cars, and these aren't the only ways in which this sector has advanced.


If you have not given too much thought to agriculture as a problem with computing aspects, take a look at the following excerpt from a report by the Kauffman Foundation on agricultural technology investment opportunities [5], which also outlines some of the startups active in this sector (p. 17, 25):


"... the most disruptive breakthroughs in AgTech may come from combining innovations in multiple areas. One particular exciting illustration of this combination is an idea known as "integrated farming systems" that will integrate genetics, physical inputs, IT sensing, and smart machinery. Through advances in software and environmental testing, farmers will be able to create custom field prescriptions for seeds, fertilizer, pest controls. Smart machinery then will carry out the prescribed treatment, all the while collecting further data that will provide feedback to the farmer."


"... Over the past five years, innovations in agriculture technology (precision ag innovations, data analytics and processing, platforms for the collection and distribution of complex data streams, and IT-driven extensions) are on the rise."


Data-driven agriculture fed by sensor networks and smart machinery may reduce the need for pesticides, decrease fertilizer consumption, limit nitrogen runoff problems [6] impacting marine ecosystems, address aquifer depletion [7], increase yields, and potentially also combat deforestation and desertification.


Incorporating computing more extensively in the food sector might improve sustainability more broadly. Consider, for example, the potential for improvements discussed by IBM [8], which can reduce the 30-40% of food production that goes to waste by improving traceability throughout the supply chain and enabling the sources of food-related infections to be isolated.


As you move closer to the plate, still further opportunities arise. Databases of the properties of foods can be analyzed to suggest [9, 10] promising flavour combinations that might result in more sustainable and affordable meals. Environmental impacts can vary significantly [11] from one type of food to another, and improving the taste and nutritional quality of low-cost food can be of particular benefit to those living in extreme poverty [12], who may choose to spend some of their scarce dollars elsewhere in search of some small pleasures and thus remain caught in a poverty trap. Malnutrition has been estimated [13] as a contributing factor in a majority of the deaths of children in the developing world, with some arguing [14] that high rates of child mortality contribute to high fertility rates thereby placing stress on the sustainability of communities.


From farm to fork opportunities remain for computing researchers to play a role in making the food system more sustainable.
 
 
[1] "Design of farm waste-driven supply side infrastructure for data centers," Sharma, Ratnesh et al, in: Proceedings of ASME 2010 4th International Conference on Energy Sustainability, Phoenix, USA, 2010, http://proceedings.asmedigitalcollection.asme.org/proceeding.aspx?articleid=1607271
[2] "Agriculture, Forestry and Other Land Use (AFOLU)", in: Climate Change 2014: Mitigation of Climate Change, Intergovernmental Panel on Climate Change, 2013, http://www.ipcc.ch/report/ar5/wg3/
[3] "Food production must double by 2050 to meet demand from world's growing population, innovative strategies needed to combat hunger, experts tell second committee", http://www.un.org/News/Press/docs/2009/gaef3242.doc.htm
[4] “Looking Forward from Two Decades of Precision Agriculture: The CropLife/Purdue University Survey”, Erickson, Bruce, in: InfoAg 2013, http://www.infoag.org/presentation/3/186/
[5] “AgTech: Challenges and Opportunities for Sustainable Growth,” Dutia, Suren, Kauffman Foundation, 2014, http://www.kauffman.org/newsroom/2014/04/agriculture-tech-advances-opportunities-for-entrepreneurs
[6] "Agricultural runoff fuels large phytoplankton blooms in vulnerable areas of the ocean," Beman, Michael J., Arrigo, Kevin R. and Matson, Pamela A. (2005), in: Nature, 434:7030 (211-214), http://dx.doi.org/10.1038/nature03370
[7] “Global depletion of groundwater resources,” Wada, Yoshihide, van Beek, Ludovicus P. H., van Kempen, Cheryl M., Reckman, Josef W. T. M., Vasak, Slavek and Bierkens, Marc F. P. (2010), in: Geophysical Research Letters, 37:20, http://dx.doi.org/10.1029/2010gl044571
[8] “Smarter Food,” IBM, http://www.ibm.com/smarterplanet/us/en/food_technology/article/food_technology.html
[9] “How A Former Google Data Guy Could Change What We Eat For Breakfast,” TechCrunch, July 2014, http://techcrunch.com/2014/07/03/how-a-former-google-data-guy-could-change-what-we-eat-for-breakfast/
[10] “We Put a Computer in Charge of Our Test Kitchen for a Day, and Here's What Happened,” Bon Appetit, June 2014, http://www.bonappetit.com/test-kitchen/inside-our-kitchen/article/chef-watson-in-the-ba-test-kitchen
[11] “Climate and Environmental Impacts,” A Meat Eater's Guide to Climate Change and Health, Environmental Working Group, 2011, http://www.ewg.org/meateatersguide/a-meat-eaters-guide-to-climate-change-health-what-you-eat-matters/climate-and-environmental-impacts/
[12] Poor Economics: A Radical Rethinking of the Way to Fight Global Poverty, Banerjee, Abhijit and Duflo, Esther, PublicAffairs, 2011.  See also http://pooreconomics.com/chapters/2-billion-hungry-people
[13] “The effects of malnutrition on child mortality in developing countries,” Pelletier, D. L. et al. (1995), in: Bulletin of the World Health Organization, 73:4 (443-448), http://libdoc.who.int/bulletin/1995/Vol73-No4/bulletin_1995_73(4)_443-448.pdf
[14] “Parental investment and the optimization of human family size,” Lawson, David W and Mace, Ruth (2011), in: Philosophical Transactions of the Royal Society B: Biological Sciences, 366:1563 (333-343), http://dx.doi.org/10.1098/rstb.2010.0297