This is the second blog post in a series that features continuing discussion with various senior staff of the Uptime Institute in response to the New York Times feature on data center energy use. The following was drawn from a discussion with Julian Kudritzki, Senior Vice President of Uptime Institute.
Many data centers are inherently wasteful due to the governing (mis)economics of cheap, high-availability computing. As long as IT reaps economic benefits divorced of true costs, then discipline in server procurement, utilization, and data center management is a ways off.
These favorable economic conditions delay motivation to change, but once dissipated, will transform our industry’s characteristics.
Hydro power west of the 100th meridian in the US is fundamentally government subsidized. (See Marc Reisner Cadillac Desert.) Federal investment in Grand Coulee generated massive amounts of inexpensive and reliable power. Today’s result is that Quincy is a desirable data center location. Power commissions seek data centers as the ideal customers of high concentrations of power. Now those bulk power deals are unavailable in the Pacific NW (scroll past login for full article).
Cost are no longer sub-3 cents, but susceptible to market fluctuations. Arguably, Quincy’s data centers are the last IT beneficiaries of a federal spend made over 50 years ago. If new data centers flock to available bulk power, which is dwindling nationally, the cost repercussions will compel a new data center mindset that looks at the IT provisioning and utilization and data center capacity deployment with a much more clinical and harsh eye. And, if we look beyond western hydro, there will be cost and lifecycle consequences of the existing carbon-intensive power generation.
The commodity server model allows for cheap and short-lived deployments. But, commodity servers are viable due to leveraged global labor pools and variable environmental regulations to dispose of the troublesome contents of a decommissioned server. Similar to hybrid cars, the sticker price does not reflect the cost of throwing it away. If foreign outsourcing of server recycling (i.e., teardown) was performed onshore at the scale of disposal, what would the total cost of a throwaway server become?
The consumer’s approach would evolve to extending the life of that server rather than replacing it. The irregular and need-based operation of diesel engine generators pales in environmental comparison to server disposal toxicity. Uptime Institute Survey shows 20% of IT departments pay their own power bill. Thus, is it safe to assume that the same percentage pay their own garbage bill?
For the enterprise, data centers and their contents are often treated as a cost center. For the IT and data center teams, the mission of uninterruptible uptime has been paramount. Thus, the prevailing management mode has been to hold nose and sign check. Budget reductions will threaten headcount in operations teams, but overall data center budgets continue to grow.
The unintended result of the favorable economic conditions is an unrealistic and unsustainable end user mindset of all IT functions available all the time. Many apps whose business value is not of the highest order have luxuriated in a ‘buy new now’ servers and enterprise-grade data center platform (i.e., power, cooling, monitoring, and automation infrastructure). Forward-thinking enterprises have been analyzing and distributing applications to match data-center infrastructure-level support to business value. But, these leading companies do not indicate industry prevalent behavior.
Debating whether server utilization is 7% or 12% or even 20% is a distraction. The issue is the economic factors that allow those low numbers to be perpetuated. There is a move afoot to compress IT, such as virtual server instances. But, the fact that this is an emergent trend shows how far we must travel and so fast.
OUTCOME: A BOON FOR OUTSOURCERS
A disruption in the economic conditions that IT has been enjoying will compel a new level of discipline and consequence in IT decision making. And, outsource alternatives will rush to propose a host of solutions to this economic crisis.
Many of these options will be so complex, or have such efficiencies of scale, that more enterprises will continue to divest themselves of data center facilities or entire IT assets.
A 2012 Uptime Institute survey of global owners revealed that 85% utilize colocation or hosting. Yet 54% had no confidence in their ability to compare outsourcing alternatives dependably. This is to the detriment of both enterprises and providers, as it calls into question the basis and viability of such commitments.
Previous IT decisions were based upon a narrower set of competitive offerings. Currently, the vast majority of enterprises deploy a hybrid computing environment. The decision is no longer binary—whether or not to outsource. It’s multi-faceted—how many and how much of each alternative to deploy.
Uptime Institute has been intensively developing the FORCSS methodology to weigh deployment alternatives with a full look at the major benefits and constraints of these options. FORCSS is the theme of Uptime Institute Symposium 2013. For more information on FORCSS click here.
Find the previous post in this conversation here. Photo of Grand Coulee Dam by Chris M on Flickr