Cost, scalability and instant access are all reasons for companies to embrace cloud computing services, but the argument that it’s a much greener option warrants further examination. Cloud computing can deliver a more efficient use of computing power. But according to a report that will be released later this month from University of Melbourne researcher Rod Tucker and his team, in some cases it consumes more energy than traditional in-office computing.
Simply put, data traveling to and from the cloud to the user’s device (often times a computer) still requires energy. Most researchers in the past have overlooked this fact in lieu of focusing solely on the energy consumed within the data center. But while the energy required to transport data from the cloud and to each device is nominal on an individual level, across thousands and millions of devices it can add up; in certain cases, transporting the data actually overtakes any efficiencies gained by putting applications in the cloud.
Applications – Not All Clouds are Created Equal
Cloud computing can indeed save energy when it leads to the consolidation of servers. But looking at three different applications of cloud computing — storage, software and processing — it’s clear that the energy efficiency savings are negated in a variety of scenarios.
One such instance is using cloud computing for storing data — the kind of service that companies like Iron Mountain and Mozy provide. Tucker found that at low download rates, cloud-based storage is more energy efficient. But when the number of downloaded and accessed files becomes larger, those energy efficiency gains are erased, because at a rate of more than one download per hour for a public cloud storage service, over 80 percent of the power used goes to transporting that data.
Similar issues arise when companies use the cloud for software-as-a-service (SaaS), like the kind Salesforce provides. The majority of enterprise software companies utilize the SaaS delivery model today, and Tucker’s research found that cloud-based SAAS used in high volume can be more energy efficient. But as the rate of the frames displayed on the end-user monitor increases (refresh rates), the percentage of power consumed in transport increases and makes the delivery model a less green option.
When it comes to using the cloud for processing — Tucker uses video encoding as an example in his report — managing the energy efficiency equation becomes even more complex; there is a lot of variation that can happen in processing. Factors depend on how modern (think efficient) the user’s computer is, how efficiently the computer does calculations and how often the end-user accesses the cloud processor. Tucker found that if the user is encoding less than four videos per week, the most energy-efficient option is to use a basic computer and a cloud processing service, as most of the power is consumed in the computer. But energy for transporting data adds up; if the user processes more than four encodings for a public cloud and eight for a private cloud, using a mid-range PC (and not cloud computing) is a more energy efficient option.
And to make another distinction, not all clouds are the same when it comes to how they consume energy. There are private clouds, which run cloud-computing services over company-owned and controlled networks. These tend to use less energy when transporting data because the network is owned and controlled. Public clouds route traffic around the Internet, and tend to use more energy as a result.
See Tucker’s handy table for a one-word break-down of the energy consuming transport nuances between private and public clouds and applications from software-as-a service, to storage-as-a-service, to processing-as-a-service.
Putting the Cloud to Good Use
Tucker’s research can be put to good use in helping companies decide when cloud computing services can be more energy efficient. It could also potentially help existing cloud computing operators make their services more energy efficient. Tucker says, “if properly managed, cloud computing can potentially lead to overall energy savings.” In particular, if data center operators and corporations upgrade their equipment to the most energy efficient possible, the savings can be a lot greater.
For example, software-as-a-service in the cloud and cloud-based processing can be much more energy efficient if both the end user devices (laptops, computers, etc.) and networking gear (like routers) are upgraded to modern low-power versions. On the other hand, Tucker says that companies using storage-as-a-service in the cloud won’t benefit all that much from upgrading to the latest equipment on either side of the network, because most of the power for any equipment comes from transporting the data.
The report’s findings also complicate newly discovered green marketing claims of firms that create or utilize cloud computing like content delivery network Akamai. Last year Bobby Blumofe, Akamai’s senior vice president of network operations, told me that because Akamai can dynamically balance customers’ traffic loads across its infrastructure, it can run its servers at two to three times higher utilization than its customers could do in-house. Better utilization of equipment means Akamai can use two to four times less server equipment than its customers would use without its help, which translates into energy saved. According to Tucker’s research, this is the ideal situation.
Perhaps most importantly, companies looking to reap any energy benefits from cloud computing should turn to virtualization first and foremost. VMware built a business off of virtualizing servers, before being greener was in fashion. Perhaps instead of emphasizing that cloud computing can be greener, companies should just cut to the chase and point to the energy efficiencies of virtualization.
All of this research also delivers another conclusion: Much more research needs to be done on energy efficiency and networking infrastructure. Tucker’s report is basically the first comprehensive analysis of the energy efficiency potentials of cloud computing; making more data like this available to companies will allow them to make even wiser choices about turning to cloud computing, as well as where to place the dollars for infrastructure upgrades.
Here are five lessons distilled from Tucker’s report on how to make cloud computing greener for your company:
- Embrace virtualization. Always. Reducing unnecessary infrastructure will almost always reduce energy consumption.
- Upgrade to modern, energy efficient PCs and laptops if your company is doing a significant amount of processing without the cloud. But if you’re going to depend on cloud for processing, save the investment on the company energy efficient computers.
- Know your company’s data load. If your company plans to store a lot of data frequently uploaded to the cloud, you’re not going to be able to claim the green angle to cloud computing. Don’t even try.
- Invest in more energy-efficient pipes. Internet infrastructure companies, from Alcatel Lucent to Cisco to the phone companies, will have to increasingly invest in more energy efficient pipes for data transport. Energy efficiency was never a problem in the days of the construction of the Internet and mobile broadband, but clearly it’s becoming more of a concern (on company budget and environment).
- Crunch your own data. If you want to make a claim about how energy efficient your cloud computing based service is, you’re gonna have to measure it yourself.
