Thursday, August 21, 2008

Jevons’ Paradox - beware the unintended consequences of driving data center energy efficiencies.

In a recent article Christian Belady[1] surfaced the Jevons’ Paradox as a caution to the enthusiasm that is sweeping the IT world when discussions focus on energy efficiency projects, programs or technologies. There is a growing expectation that technology driven conservation will solve the looming data center energy crises however if Jevons’ Paradox is to be believed conservation is only part of the answer. Let me be very clear before I continue with this dialog, energy conservation and efficiency efforts are good things and are to be applauded. I am not arguing against the wisdom of saving energy, eliminating waste or encouraging vendors to develop energy efficient technologies, quite the opposite, prudent oversight of corporate resources while delivering against corporate expectations is obligatory for those responsible for data center/IT performance.

The Jevons’ Paradox is an interesting observation published in 1874 by William Stanley Jevons in his book, The Coal Question. Jevons noted that, “the greater the energy efficiency, while in the short run will produce energy savings, may in the long run result in higher energy use”. His paradox is that “it is a confusion of ideas to suppose that the economical use of fuel (energy) is equivalent to diminished consumption. The very contrary is the truth” Belady puts it a bit more succinctly “As technological improvements increase the efficiency with which a resource (energy) is used, total consumption of that resource may increase rather than decrease”

To put this observation into perspective and relate it to the reality of IT and the data center you only have to look at what has happened in the storage industry. Technology has dramatically improved the bit/sq inch efficiency by driving areal densities (bits/sq inch) from 10Mb in the 1980’s, to 100Gb today and to a projected 1Pb by 2025[2]. What this means for the end user is a dramatic reduction in device acquisition cost ($/MB), reduced lifetime cost of the storage device (reliability, management, floor space, energy) and significant improvement to the performance/cost ratio (Moore’s Law). However has this evolutionary cost reduction resulted in a corresponding decline in the need for storage solutions or of the actual dollars spent on storage? The answer is a resounding NO! Cheap (historically speaking) storage has enabled explosive data growth fueled by the current propensity of enterprises to want to permanently store massive amounts of accumulating data, effectively the corporate digital memory, ready to have its value extracted by data mining applications or to be available to satisfy the pervasive needs of regulatory compliance. Lower storage costs facilitates the purchase of more storage.

Cheaper, higher performing data storage has enabled many business applications which in turn drive the creation of more data, needing more storage, and so the cycle continues. This is illustrates what is called the Jevons’ Paradox, the rebound effect.

OK, back to the topic. Energy conservation whether driven through project, programs or technologies will free up data center power. One scenario would say that this will lower data center operating cost to the benefit of bottom line. Remember the rebound effect.
The second and more realistic scenario says that any power savings realized will be applied to the support additional equipment and to drive greater compute performance. In short, cost is not the primary architect of what has been called by the Uptime Institute as the “ silent crises in the data center” it is power availability and the impact power constraints will have on data center performance. This does not say that energy cost is not an issue, it is, but availability is number one followed closely by cost. After all IT, who is the custodian the corporate digital memory, runs on electricity and if electricity is constrained then IT is constrained and corporate productivity will be constrained. Who wants to be at the bottom of that pile? Why do you think Google, Yahoo and the NSA have data centers near hydroelectric power plants in Oregon? Why is HDS near the geothermal energy resource in Greenland? Cost is part of the answer but securing a reliable source of large amounts of energy to sustain future growth is the primary motivation.

In a recent WSJ article it was pointed out that from a macro level in the US the GDP/unit of energy is twice the level achieved in the 50’s while energy usage has increased threefold. This means that energy and growth are directly connected, a lesson the developing nations already understand and so it is in the data center, energy availability fuels the productivity needed for sustained, long term growth.

AAAH!!! I hear the critical voice say, virtualization has solved the problem! Server and storage consolidation is the answer! Nope, not so fast. Virtualization will most certainly help drive greater efficiencies and the short term gains will be enjoyed. However, if the Uptime Institute is to be believed, all the servers that can be virtualized will be by 2011, similar scenario for storage.

This brings me to the primary objective of this post which is to suggest that while we should enthusiastically do all that we can to conserve energy and while we may see a dip in power consumption the reality is that energy is needed to fuel growth. We can expect that following any meaningful conservation efforts energy demand will rebound to original consumption levels and beyond. This is a contrarian opinion to that expressed in the EPA 2007 report to congress which suggests that technology driven conservation will reverse the upward trend for data center power. I hypothesis that we should be more concerned about making sure there is an adequate supply of (affordable) energy rather than looking myopically at securing short term efficiency gains.

So, if you accept my reasoning, you will also accept my conclusion that the impact of conservation efforts (IT equipment) will produce relief to the energy issue but it will be temporary. The much larger, and significantly more critical issue is the need to increase the capacity of the current power grid and to drive down the cost of energy. Get on the phone with your legislative representative irrespective of political brand and demand that the US focuses on increasing the capacity power grid and let us not be confused with notion that conservation itself will solve the IT energy challenge. Remember IT runs on electricity!

As always I look forward to comments.
[1] “Does energy Efficiency in the Data Center give us what we need?”; Mission Critical, Spring 2008: Christian Belady, Power and Cooling Architect, Microsoft;
[2] Fred Moore; Horison Information Strategies

No comments: