Issue No.01 - January/February (2008 vol.10)
Published by the IEEE Computer Society
Joseph Williams , Microsoft
Lewis Curtis , Microsoft
DOI Bookmark: http://doi.ieeecomputersociety.org/10.1109/MITP.2008.9
Although currently there's a lot more hype than science or engineering behind green computing, there's also great promise behind achieving the goals that are being articulated.
Getting beyond the veil of marketing to realistic goals will be challenging for IT professionals not trained to understand the complexities and interdependencies of how products, architectures, and operational processes roll up to a single green scorecard. A whole raft of consultants has sprung into action to sell everything from green computing audits to a greener way of doing business, while IT vendors are positioning products from green supercomputers to green recycled mouse pads.
The rush to green has resulted in speculation ranging from Web server farms patterned after energy-efficient dancing bees ( www.msnbc.msn.com/id/22266034) to the creation of a new chief energy officer role who would drive green computing initiatives ( www.onstor.com/pr_09_18_2007.php). It has also spawned a number of interesting technical debates, such as whether a 35-watt thinclient workstation with a 25-watt sleep mode is more energy efficient than a 60-watt PC with a 4-watt sleep mode.
Making sense of it all is going to take years. Staying informed while getting involved is, as always, a prudent way to approach a trendy IT topic.
In this Issue
This issue of IT Professional presents some interesting views about green computing Each of the authors has a lot of passion for the topic, and their articles will help you frame your thinking and challenge some of the conventional wisdom.
In "Green Supercomputing Comes of Age," Wu-chun Feng, Xizhou Feng, and Rong Ge discuss how power and cooling issues in supercomputing have evolved from a secondary concern to a primary design constraint. Meanwhile, San Murugesan, in "Harnessing Green IT: Principles and Practices," argues not only that IT businesses can gain a competitive edge by adopting green IT practices, but also that the IT sector as a whole has a responsibility to help create a more sustainable environment.
"Making IT a Positive Force in Environmental Change," is a Perspectives article by Jesse S. Aronson that addresses ways to mitigate IT's negative affects on the environment. Interestingly, though, he also outlines the many ways that technology helps reduce energy use in other industries.
In IT Pro's new IT Pro/Con department, Barry Shevlin notes the environmental benefits of recycling by selling and using refurbished equipment in the article, "When and How to Use Refurbished Equipment for IT Needs." And, in the Trends department, Jan Krikke covers the booming market of e-waste recycling in "Recycling e-Waste: The Sky Is the Limit."
Alongside the other articles in this special issue, we'd like to briefly address the aspect of green computing that reflects our own passion—namely, what an architectural approach to the topic might look like. This is not a definitive statement of what a green architecture would look like—rather, it describes what constructing that architectural view will require.
"Green" isn't all that new, of course. Energy management has always been an important topic for laptop manufacturers, who have been working on this for years as they struggled to manage weight versus battery life issues (see Figure 1, which is a chart we've recreated from Intel's white paper "The Battery Life Challenge: Balancing Performance and Power," http://whitepapers.silicon.com/0,39024759,60295305p-39000684q,00.htm). The powerful new laptop with the terrifically beautiful video display might require a more expensive battery technology if the laptop's weight is to be maintained as well as its battery life. Longer battery life always comes with a price—typically a guns-or-butter trade-off (if you remember Economics 101) between heavier batteries or more expensive battery technologies for every additional minute of battery life. Perhaps you can keep the beautiful video display and the cheaper battery, but it comes at the cost of a lower-performing but more energy-efficient CPU and hard drive. Typically, even when weight or costs come down, more energy-hungry components squander what energy and cost savings could have been achieved.
As with all architecture, the fundamental principle is to understand and rationalize the inevitable trade-offs.
An Architectural Approach to Green Computing
There's no simple path to green computing, but there are some low-hanging fruit. You can spin the dial on some straightforward actions, such as orienting racks of servers in a data center to exhaust their heat in a uniform direction, thus reducing overall cooling costs. You can also implement Energy Star guidelines ( www.energystar.gov/index.cfm?c=prod_development.server_efficiency) for energy efficiencies in the data center. But these are point solutions. A comprehensive plan for achieving green computing really does require an architectural approach.
Because of IT's ecosystem complexities—ranging from the data center to client computing and from customer impacts to business impacts—some investment and process decisions almost always involve trade-offs, or, if you aren't architecturally proactive in your thinking, a tapestry of unintended consequences. You won't find a silver bullet or a single vendor solution that will magically make anyone "green." In fact, given all the interdependencies and complexities, it isn't entirely obvious yet what a green outcome would even look like.
The architecture in 3D
Figure 2 provides one view of green computing's multidimensional architectural complexities. The Green Architecture Cube calls out the interaction of technology, customer impact, and metrics. The Cube is not all encompassing, but it's designed to facilitate a general discussion of the green ecosystem while also providing a planning roadmap of sorts.
The technology dimension of the cube doesn't address only the data center. It's a collection of the green impacts from all of the company's IT facilities, communications infrastructure, management platform, business and operations processes, and client platform. In short, an organization must consider the entire technology ecosystem when reviewing green initiatives.
For example, a company's decision to lower overall corporate carbon emissions by reducing travel would put additional emphasis on collaboration tools as a substitute for travel. This might require an added rack of servers, which will place extra strain on the data center facility and its resources. The rack would definitely put more of a load on the communications infrastructure, and that in itself might entail deploying more routers, switches, load balancers, and other energy-consuming devices that also must be powered and cooled. These servers and devices will require management, which will put load on the management servers and additional load on the network. They might also result in an increase in headcount to manage the additional data center load. The new employee would drive to work, consume office space and its related resources, and require an energy-consuming computer.
Substituting online collaboration for travel may well lead to operational changes necessary to provide better, more flexible, and more robust collaboration solutions. In turn, the reduction in travel will require material changes in the way that business is conducted. At the end-user level, more emphasis on collaboration might require investing in new applications, peripherals (such as two-way video cameras), and an enhanced user-support infrastructure.
These technology areas, in turn, variously but collectively impact stakeholders in unique ways. Customers who are now getting more digital collaboration might need to upgrade their own infrastructures to handle the load. Employees who are no longer driving to airports or sitting on planes but who are also losing some measure of personal interaction will have green as well as personal impacts. Partners might or might not adjust to the travel-less interactions, but they certainly would need to tune their own technology infrastructures to keep pace.
The general public—whether consumers buying at a big-box store who no longer have product associates there in person to help with promotions, or college students who are interviewing at this now reduced-travel company—will have different reactions and impacts from this policy. In turn, these impacts could have material consequences to the way the company runs IT and its business operations. As regulatory interest increases, particularly in Europe, the impacts to government stakeholders must be examined.
An ecosystem map
Another view of green computing's architectural complexities flattens out the Green Architecture Cube into more of an ecosystem map, as depicted in a simple example in Figure 3. Here you can more clearly see the constituent components that collectively come to bear in a green architecture. In a large IT operation, this tree could conceivably span hundreds if not thousands of branches, particularly if it includes multiple business units and data centers located around the world.
You can also get a clearer view of how deeply the architecture for green computing cuts across so many functional roles in an IT operation. This gives you a better feel for why there isn't going to be a silver (or should we say "green"?) bullet in any sizeable IT operation. Another benefit of this view is that it shows the landscape across which a green audit should examine if it's to be truly meaningful.
Figure 3 also shows the complex relationships and stakeholders that must be considered. An architectural approach to green computing accounts for these relationships and stakeholders and provides a framework for rationalizing how IT can more effectively address the challenges of going green.
Measures of Success
Ultimately, all of these interactions must be measured in some way to ascertain whether, in fact, a particular policy will result in the desired outcome. Unfortunately, even this requirement is complicated because the metrics vary depending on the stakeholder's view. Business value metrics, such as total cost of ownership and return on investment, are a given, but those must be considered against the carbon footprint, energy consumption, and pollution-generating metrics, as well as against the value or burden of regulatory compliance or access to government programs such as tax credits, plus the brand value of being able to show commitment to sustainable business operations.
Green computing has no widely accepted metrics, so you must be careful to compare apples to apples when taking an architectural view. The Uptime Institute ( www.uptimeinstitute.org) has published definitions for some quantitative metrics for measuring a data center's green quotient (for example, SI-POM to measure the site infrastructure power overhead multiplier, which is the ratio of at-the-plug power consumption to total-site power consumption). These metrics are similar in some ways to how Greenpeace built its electronics metrics ( www.greenpeace.org/electronics) and to the Power Usage Effectiveness and Data Center Efficiency metrics proposed by the GreenGrid ( www.thegreengrid.org). But they aren't the same metrics—and used improperly, they can lead to contradictory conclusions about key green performance indicators. At some level, it's likely that government or industry authorities will step in with more widely accepted metrics that will help solve this problem. But again, we must be careful to take the whole picture into account when measuring green success or progress.
Using an analogy from security, in the past 15 years IT has learned that purchasing a firewall doesn't necessarily make an IT environment secure. It takes an entire security architecture to provide a vision against which secure computing can be executed. The same is true for green computing. Point solutions will help, but really addressing the issue takes an architectural view.
The authors' views represented in this article do not represent the views of Microsoft Corporation.
Joseph Williams is the CTO of WW Enterprise Sales at Microsoft. Contact him at firstname.lastname@example.org.
Lewis Curtis is an architect for developer platform evangelism at Microsoft. Contact him at email@example.com.