According to data released Tuesday as part of Cisco’s Global Cloud Index, IP traffic over data center networks will reach 4.8 zettabytes a year by 2015, and cloud computing will account for one-third of it, or 1.6 zettabytes. The report itself is a refreshing reminder that while public Internet traffic gets most of the attention, data centers are working hard behind the scenes to make the web run. By 2015, data center networks will handle about five times the amount of data that will travel across the Internet.
To wrap one’s head around the disparity in traffic, one need only look at the life of a single email message. As Cisco’s Shruti Jain explained to me, a single 1MB message sent to a handful of people can account for about 50MB of data after each email system stores, replicates, scans and otherwise deals with the message. Data centers somewhere, whether within individual organizations or those operated by service providers such as Google, handle all those tasks.
Of course, we’re using the Internet for a lot more than sending emails these days, which is why data center traffic volume is growing so fast. According to Cisco, data center traffic already was at 1.1 zettabytes in 2010, and will grow at a compound annual growth rate of 33 percent through 2015. Of that, traffic coming from cloud services will grow at a CAGR of 66 percent, from 121 exabytes in 2010 to 133 exabytes a month in 2015.
For the purpose of the study, Cisco defined cloud services as having the following characteristics:
- Ubiquitous access. The application or service can be accessed anytime, anywhere from multiple devices.
- Persistent personalization. The application or service can apply basic user credentials and customized preferences or settings.
- Virtual capacity/processing. The application/service can store or process data beyond the capacity of (and separate from) the connected end-user device point.
- Flexible delivery. The application or service offers scalable provisioning, usage-based pricing, and can be delivered on-demand.
For cloud computing specifically, the Global Cloud Index marks 2014 as a magic year: It’s the year cloud-based workloads will surpass traditional-data-center-based workloads in total volume.
Consumers rule in the data center, too
Probably not surprisingly, Cisco expects consumer-based workloads to dominate the traffic on data center networks. Cisco’s Doug Webster told me although business traffic actually will grow in volume from 259 exabytes in 2010 to 744 exabytes in 2015, that threefold growth pales in comparison to consumer traffic growth, which will grow nearly 5x by 2015.
Webster said it’s workloads such as streaming video and other data-intensive activities that will power the surge in consumer traffic. In fact, he said, during “Internet primetime” in 2015 — the hours when the most people are online — data centers will be 2.5 times busier than they are on average. Internet traffic increases about 25 to 50 percent during those same hours.
The study notes the increase in consumer traffic levels within data centers is largely attributable to cloud-based applications. Between 2000 and 2008, it states, peer-to-peer traffic dominated consumer Internet traffic, which didn’t put a strain on data centers. By 2015, however, 37 percent of consumer data center traffic will come from cloud applications, up from 14 percent in 2010.
Who’s ready for the cloud?
One final aspect of Cisco’s Global Cloud Index is an assessment of how ready certain geographic regions are to support various types of cloud-based applications. While all the regions — Latin America, the Middle East, Asia-Pacific, North America, Central and Eastern Europe, and Western Europe — are capable of supporting basic applications such as email, text messaging and social networking, only the latter five can handle basic applications such as HD video streaming, basic gaming and basic video chat.
Webster said that no regions, as a whole, can support advanced cloud applications such as advanced gaming, video chat and video conferencing, but that finding is somewhat misleading. Almost every region has average download and upload speeds capable of handling those applications, but latency is the major stumbling block. And, as both Webster and the study note, there are outliers in every region (e.g., Japan and South Korea in APAC, and the United Arab Emirates in the Middle East) that are much more ready than other countries in their areas.
One interesting finding is that while North America lags near the middle in average upload and download speeds, it leads in both fixed broadband (27 percent) and mobile broadband (54 percent) penetration rates.
Overall, Cisco’s study is valuable not so much because it tells us something we didn’t know — anyone following the web infrastructure space has seen data center growth skyrocket over the past couple years — but because it quantifies just how much traffic data centers are dealing with. The good news for the growing number of individuals and businesses that rely on cloud-based applications is that although the Internet itself might be under a great deal of strain, many service providers have proven very willing to invest in resources and architectures to ensure their data centers can handle whatever loads they’re asked to deal with.