Cisco: Cloud-Native Apps Will Dominate By 2020

CIOs Face Changes, Volatility In 2017, Forrester Predicts

By 2020, the internet of things (IoT), analytics, and database applications will make up 22% of all business workloads in cloud data centers, an increase from the 20% businesses saw in 2015. These big data applications will most likely be "cloud-native," or designed to run in the cloud, and the growing use of them will be at the center of enterprises' movement into the cloud.

That move is about to accelerate.

However, even as the enterprise moves into the cloud, its claim on the world's total cloud computing resource will shrink compared the role of the consumer, according to the Cisco Global Cloud Index released Nov. 10.

In 2015, enterprises accounted for 79% of the workloads in either public or private clouds. By 2020, that share will drop to 72%, offset by the growing consumer use of the cloud. In 2015, consumer workloads accounted for 21% of public cloud use.

By 2020, they will account for 28%. While enterprise use of the cloud appears to be growing at an annual compound percentage rate of 19% through 2020, consumer use is growing faster -- at a CPR of 28%.

Consumer applications taking up the most bandwidth in the cloud data centers by 2020 are projected to be: video streaming workloads, at 34% (versus 29% in 2015); social networking, at 24% (versus 20% in 2015); and search workloads, at 15% (versus 17% in 2015).

These were key conclusions from the annual Cisco Global Cloud Index, which is now in its sixth year.

Hyperscale Ramps Up

For the first time, Cisco was able to analyze the categories of enterprise workloads in the data it was collecting. It also put a number to how many cloud-scale data centers have been built so far by the likes of Facebook, Google, Amazon, Apple, IBM SoftLayer, Microsoft, Rackspace, and other service providers.

That number was 257 in 2015, and it's slated to grow to 485 by 2020.

That's a reference to the hyperscale data centers that web services companies have been building, such as Facebook's Prineville, Ore., facility or Microsoft's data center outside of Chicago.

Enterprises also build large data centers, but a hyperscale data center is loosely defined by Thomas Barnett, Cisco's director of research for the Global Cloud Index, as meaning a data center that is associated with "billions of dollars of revenue" from Infrastructure-as-a-Service (IaaS), Software-as-a-Service (SaaS), or other cloud services.

Synergy Research, which produces data used in the index, detailed in a footnote that hyperscale means a facility that produces $1 billion in IaaS or Platform-as-a-Service (PaaS), $2 billion in SaaS, $4 billion in an internet service like search, or $8 billion in payment processing.

Facebook's Prineville site, a Google facility powering search, and the Amazon Web Services complex in Ashburn, Va., would all qualify.

Another way of putting it is that is hyperscale data centers contain hundreds of thousands of servers. Facebook's first hyperscale data center covered 330,000 square feet. Nearby was an Apple data center covering 338,000 square feet. Microsoft's Chicago facility was built to host up to 300,000 servers.

"Hyperscale data center traffic is projected to quintuple over the next five years. These infrastructures will account for 47 percent of total data center installed servers and support 53 percent of all data center traffic by 2020," Doug Webster, vice president of service provider marketing at Cisco, noted in the announcement of the index.

The hyperscale data centers will account for 47% of all data center servers installed as they continue to be built out through 2020. They will support 53% of all data center traffic by 2020.

Overall, cloud traffic will amount to 14.1 zettabytes, or 14.1 trillion gigabytes -- up from 3.9ZB in 2015. That represents a 3.7-fold increase in overall cloud traffic.

[Want to look inside a hyperscale data center? See the photos Facebook's Futuristic Data Center: Inside Tour.]

This traffic is being moved more efficiently, thanks to software-defined networking and network function virtualization. Over the next five years, 60% of the hyperscale data centers will have deployed SDN and NFV, the index reported.

Also by 2020, 44% of all traffic will be over the virtualized networks.

Public vs. Private

Workload density in the cloud keeps increasing. In 2015, it was 73 workloads per physical server. By 2020, it will be 11.9 workloads. Enterprise data centers, in contrast, hosted 2.2 workloads per server in 2015 and will host 3.5 by 2020.

The public cloud is growing faster than the private cloud, usually represented by cloud-like operations in the enterprise data center. The public cloud hosted 49%, or 66.3 million workloads in 2015. It will take on 68%, or 298 million in 2020, the index projected.

Likewise, the cloud, public and private, will dominate computing by 2020, accounting for 92% of workloads, with 8% remaining in traditional data centers, according to the index, compiled by Cisco authors Thomas Barnett, director of service provider forecasts and trends, and senior analysts Taru Khurana, Usha Andra, Arielle Sumits, and Shruti Jain.