How do you measure the cloud's environmental impact?

How do you measure the cloud's environmental impact?

How green is the cloud? Reading these days, you’d think it was the next big green thing. In recent weeks, for example, we’ve covered companies bringing products to the cloud, research findings on how the cloud is a “game changer” for energy managers, cloud-based tools for designers, even ethical concerns related to cloud computing.

As my colleague Matthew Wheeland noted recently: “Everyone is talking about cloud computing as the future of IT, or the future of business, or the future of commerce, or all of the above.”

How green is it? Conventional wisdom is that cloud computing — the delivery of computing as a service rather than a product, in which computers access shared resources, software, and information, instead of each computer having its own — is a highly eco-efficient medium. In addition to dramatically reducing hardware and software, the cloud maximizes the utilization of all computing resources, at least in theory.

But calculating the cloud’s environmental benefits has been an elusive calculation, until now. — whose software-as-a-service business makes it one of the pioneers of cloud computing — recently proffered a metric for measuring the cloud’s environmental performance.

“We’re proposing carbon per transaction as the cloud performance metric that can best guide customer choice as the cloud becomes the dominant IT model,” Sue Amar, Sustainability Officer at, wrote recently on the company’s blog.

Amar explained the difficulty of comparing the environmental impact of different technologies. “For example, in the enterprise software space alone, technology can be delivered on-premise (single tenant architecture), through a private cloud (virtualized single tenant architecture) or via multi-tenant cloud computing like”

Last year, the company commissioned a white paper from WSP Environment & Energy (download – pdf) “to compare the energy use and carbon footprint of its cloud platform and services against equivalent on-premises and virtualized data center deployments.” The study determined that Salesforce’s multi-tenant model was 95% less carbon-intensive than on-premise software, with 64% less CO2 footprint than even the latest “private cloud” installations. Wrote WSP:

Dedicated infrastructure associated with on-premises deployments typically operate at very low utilization levels and do not benefit from the key features of public cloud computing; multi-tenancy and elastic provisioning which offer scalability (both up and down) and higher utilization rates. While a private cloud does benefit from elastic provisioning and higher utilization rates, it is clear that multi-tenancy is a critical driver of carbon efficiency.

Photo of businessman standing alone at the field by Helder Almeida via Shutterstock.

WSP measured one technology to another. But how do you compare environmental impacts from one service to another? Multi-tenant cloud providers have yet to standardize on the best way to measure impact,” says Amar. “Google, Facebook and others currently consider everything from data center efficiency to the renewable energy investments and the percentage of carbon emissions they offset.”

Boiling it all down to carbon emissions per transaction levels the playing field, she says:

While different technology models will yield different transactions, all data centers exist to process some kind of customer transactions. Too many measures of IT ‘greenness’ ignore this. They treat every step in an IT operation as if it were equally useful, but running a data center this way is like driving a car with a tachometer but no speedometer. It’s measuring the efficiency of the machine, but ignoring the efficiency of the process. Cloud models radically streamline core IT operations, as well as performing those operations more efficiently. Measuring the carbon output per transaction therefore offers a far more illuminating comparison than every other measure in common use today.

Using WSP’s methodology, Salesforce found that despite processing 63 percent more transactions in 2011 over 2010—including Chatter posts, Radian6 searches, and edits to contact records—the average amount of carbon produced per transaction dropped by 20 percent.

“Our ability to decrease our carbon per transaction metric is a result of our relentless commitment to energy-efficient architecture,” says Amar. “Last year, our team eliminated idle server capacity and invested in cutting-edge server infrastructure to improve the efficiency of our entire service.” Moreover, she said, “We greatly improved server capacity and utilization, as well as leveraged rapid increases in semiconductor speeds by refreshing machines and adding server pods.”

Amar believes that by measuring results rather than merely monitoring activity, the new carbon-per-transaction metric will make it easier for organizations to make smarter choices about their IT investments.

And, presumably, for their customers to compare one service to another.