In-Depth
Is Green IT Falling Drastically Short?
Why the current crop of Green IT initiatives isn't green enough
Green IT. Everyone wants it or is paying lip service about wanting it.
Three of the biggest computing OEMs, for example -- Hewlett-Packard Co. (HP), IBM Corp., and Sun Microsystems Inc. -- have made noise about the greening of their respective server platforms, with Big Blue even ponying up $1 billion to emphasize its eco-credibility. These days, HP, IBM, Sun, and other vendors push server, storage, and networking products that are denser, cooler, and require less power (relative to their information-processing capabilities) than systems they marketed just three years ago.
No vendor can afford to rest on its laurels, according to veteran industry watcher Wayne Kernochan, a senior IT advisor with consultancy Illuminata, who says most of today's Green IT offerings suffer from a crippling shortcoming: they aren't green enough.
The issue, he notes, is that much of what the big hardware OEMs are doing is beside the point.
"It isn't because what they are doing isn't striking. Announcements from the likes of AMD, EMC, HP, IBM, Intel, and Sun about the potential for 50 percent, 60 percent, and even 90 percent reductions in the energy expended by various systems and data centers, even while computing power is increasing, certainly indicate that improvements in computer technology's ability to reduce energy output may be moving even faster than Moore's Law right now," Kernochan concedes. "But the real requirements for reductions in energy output and carbon footprint to combat global warming -- the main focus of green computing initiatives -- are tougher still."
Kernochan points out that while computing designs have definitely become more efficient, they're also denser and more powerful. One upshot is that while today's quad-core processors might be significantly more efficient than their single-core predecessors, they also require significantly more power. In other words, while the individual cores of a quad-core chip might be drastically more efficient than a single-core design, the quad-core chip in toto requires substantially more power.
"We have been proceeding as if today's market-driven improvements in energy efficiency are enough, even though their result is still increasing total energy emissions," he continues. "Thus, an effective requirement is reducing total global energy emissions every year, eventually reaching a 'steady state' with minimal additional effects on the environment. This, in turn, means reducing energy emissions every year in every area with a significant energy impact."
That's not happening on the IT front, Kernochan points out. If anything, in fact, IT-related power requirements are drastically increasing.
"The bad old days around 2000, when we were hearing alarmist stories about how Web server farms used up a major part of California's energy, appear to have returned," he continues, citing an estimate (which he attributes to IBM) that data centers used 183 billion kilowatt hours of electricity last year. "In other words, the need for green computing is not just that data centers are running too hot; the overall computing-industry use of electricity is now a significant proportion of the world's energy usage."
Just how much will IT-related power demands grow over time? Kernochan cites a factoid he says he picked up at a recent EMC Corp. conference.
"[G]lobal storage size will continue to increase by 60 percent year to year over the next five years, reaching a zetabyte in 2011," he says. "The implication is that if nothing else changes, energy emissions from computing will begin to increase close to 60 percent year to year as storage energy usage begins to dominate overall computing energy usage."
In this respect, most Green IT efforts fall well short of the mark, Kernochan argues. "Things like taking a holistic view of data center energy usage, using virtualization to pack more computing power into a single machine, or figuring out how to reduce per-chip and per-disk energy requirements by better designs and better venting of heat are not only the easiest targets but also one-shots," he indicates. [T]hey do not continue to improve at the same rate over time. Once a computing unit is fully utilized via virtualization, only minor improvements in energy savings via better virtualization are typically possible."
Kernochan himself doesn't offer any technological prescriptions -- other than that vendors should avoid the low-hanging fruit -- the easy one-offs -- and focus on the more philosophical stuff. "[Vendors] have been very effective in identifying one-shots that will take us one year [such as EMC's pledge to reduce storage energy requirements by half next year], two … or even three … years down the road," he concludes.
"What they have not done is to clearly identify the processes that will allow them to do this five, 10, [or] 20 years down the road."