Cisco Global Cloud Index: Forecast and Methodology, 20152020

White Paper
Cisco Public

Cisco Global Cloud Index:

Forecast and Methodology,
20152020

2016 Cisco and/or its affiliates. All rights reserved.

Cisco Global Cloud Index: Forecast and Methodology, 20152020

White Paper
Cisco Public

What You Will Learn

The Cisco Global Cloud Index (GCI) is an ongoing effort to forecast the growth of
global data center and cloud-based IP traffic. The forecast includes trends associated
with data center virtualization and cloud computing. This document presents the details
of the study and the methodology behind it.
Forecast Overview
Hyperscale Data Centers

Hyperscale data centers will grow from 259 in

number at the end of 2015 to 485 by 2020. They
will represent 47 percent of all installed data
center servers by 2020.
Traffic within hyperscale data centers will
quintuple by 2020. Hyperscale data centers
already account for 34 percent of total traffic
within all data centers and will account for 53
percent by 2020.

Global Data Center Traffic

Annual global data center IP traffic will reach 15.3

zettabytes (ZB) (1.3 ZB per month) by the end of
2020, up from 4.7 ZB per year (390 exabytes [EB]
per month) in 2015.
Global data center IP traffic will grow 3-fold over
the next 5 years. Overall, data center IP traffic will
grow at a compound annual growth rate (CAGR)
of 27 percent from 2015 to 2020.

Data Center Virtualization and Cloud

Computing Growth

By 2020, 92 percent of workloads will be

processed by cloud data centers; 8 percent will
be processed by traditional data centers.
Overall data center workloads will more than
double (2.6-fold) from 2015 to 2020; however,
cloud workloads will more than triple (3.2-fold)
over the same period.
The workload density (that is, workloads per
physical server) for cloud data centers was 7.3 in
2015 and will grow to 11.9 by 2020. Comparatively,
for traditional data centers, workload density was
2.2 in 2015 and will grow to 3.5 by 2020.

Global Cloud Traffic

Annual global cloud IP traffic will reach 14.1 ZB

(1.2 ZB per month) by the end of 2020, up from
3.9 ZB per year (321 EB per month) in 2015.
Global cloud IP traffic will almost quadruple
(3.7-fold) over the next 5 years. Overall, cloud
IP traffic will grow at a CAGR of 30 percent from
2015 to 2020.
Global cloud IP traffic will account for more than
92 percent of total data center traffic by 2020.

Cloud Service Delivery Models

By 2020, 74 percent of the total cloud workloads

will be software-as-a-service (SaaS) workloads,
up from 65 percent in 2015.
By 2020, 17 percent of the total cloud workloads
will be infrastructure-as-a-service (IaaS)
workloads, down from 26 percent in 2015.
By 2020, 8 percent of the total cloud workloads
will be platform-as-a-service (PaaS) workloads,
down from 9 percent in 2015.

Workloads by Application

By 2020, enterprise workloads will account for 72

percent of total data center workloads, down from
79 percent in 2015.
By 2020, consumer workloads will account for 28
percent of total data center workloads, up from
21 percent in 2015.
Within the enterprise segment, compute (29 percent
of enterprise workloads by 2020) and collaboration
(24 percent of enterprise workloads by 2020) are
the two main contributors to workload totals.

Public vs. Private Cloud

Within the consumer segment, video streaming

(34 percent of consumer workloads by 2020)
and social networking (24 percent of consumer
workloads by 2020) are the two main contributors
to workload totals.

By 2020, 32 percent of the cloud workloads will be

in private cloud data centers, down from 51 percent
in 2015 (CAGR of 15 percent from 2015 to 2020).

Within the enterprise segment, database/analytics

and IoT will be the fastest growing applications,
with 22 percent CAGR from 2015 to 2020, or
2.7-fold growth.

By 2020, 68 percent of the cloud workloads will be

in public cloud data centers, up from 49 percent in
2015 (CAGR of 35 percent from 2015 to 2020).

2016 Cisco and/or its affiliates. All rights reserved.

Cisco Global Cloud Index: Forecast and Methodology, 20152020

White Paper
Cisco Public

Within the consumer segment, social networking

(33% CAGR from 2015 to 2020) and video
streaming (32 percent CAGR from 2015 to 2020)
will be the fastest growing applications.

Data Center Storage

By 2020, data center storage installed capacity

will grow to 1.8 ZB, up from 382 EB in 2015,
nearly a 5-fold growth.
By 2020, the total global installed data storage
capacity in cloud data centers will account for 88
percent share of total DC storage capacity, up
from 64.9 percent in 2015.

Data in Data Centers, Big Data, and IoE

Globally, the data stored in data centers will

quintuple by 2020 to reach 915 EB by 2020, up
5.3-fold (a CAGR of 40%) from 171 EB in 2015.
Big data will reach 247 EB by 2020, up almost
10-fold from 25 EB in 2015. Big data alone will
represent 27 percent of data stored in data
centers by 2020, up from 15 percent in 2015.
The amount of data stored on devices will be 5
times higher than data stored in data centers, at
5.3 ZB by 2020.
Driven by the Internet of Things, the total amount
of data created (and not necessarily stored) by
any device will reach 600 ZB per year by 2020,
up from 145 ZB per year in 2015. Data created is
two orders of magnitude higher than data stored.

Consumer Cloud Storage

By 2020, 59 percent (2.3 billion) of the consumer

Internet population will use personal cloud storage,
up from 47 percent (1.3 billion users) in 2015.
Globally, consumer cloud storage traffic per
user will be 1.7 gigabytes per month by 2020,
compared to 513 megabytes per month in 2015.

Multiple-Device and -Connection Ownership

North America (7.3), followed by Western Europe

(5.5), had the highest average number of devices
or connections per user in 2015, followed by
Middle East and Africa (5.4), Latin America (4.7)
Central and Eastern Europe (4.5), and Asia Pacific
(4.5).
By 2020, North America (13.6), followed by
Western Europe (9.9), will have the highest
average number of devices or connections per
user, followed by Central and Eastern Europe
(6.2), Latin America (5.2), Middle East and Africa
(5.0), and Asia Pacific (5.0).

Regional Cloud Readiness

Network Speeds and Latency

Asia Pacific leads all regions with an average fixed

download speed of 33.9 Mbps. North America
follows with an average fixed download speed of
32.9 Mbps. Central and Eastern Europe and Asia
Pacific also lead all regions in average fixed upload
speeds with 19.3 Mbps and 19.0 Mbps, respectively.
Asia Pacific leads all regions in average fixed
network latency with 26 ms, followed by Central
and Eastern Europe with 30 ms.
Asia Pacific leads all regions with an average
mobile download speed of 18.5 Mbps. Western
Europe follows with an average mobile download
speed of 18.2 Mbps. North America and Asia
Pacific lead all regions in average mobile upload
speeds with 9.9 Mbps and 8.9 Mbps, respectively.
Western Europe and Asia Pacific lead all regions
in average mobile network latency with 57 ms
and 73 ms, respectively.

Top Seven Data Center and Cloud

Networking Trends
Over the last few years, the telecommunications
industry has seen cloud adoption evolve from an
emerging technology to an established networking
solution that is gaining widespread acceptance and
deployment. Enterprise and government organizations
are moving from test environments to placing more
of their mission-critical workloads into the cloud. For
consumers, cloud services offer ubiquitous access to
content and services, on multiple devices, delivered
to almost anywhere network users are located.
The following sections identify seven important
trends in data center and cloud networking that
are accelerating traffic growth, changing the enduser experience, and placing new requirements
and demands on data center and cloud-based
infrastructures.
1. Growth of Global Data Center Relevance
and Traffic
-- Growth of Hyperscale Data Centers
-- Global Data Center IP Traffic: Three-Fold
Increase by 2020
-- Data Center Traffic Destinations: Most Traffic
Remains Within the Data Center
-- Global Data Center and Cloud IP Traffic Growth
-- SDN/NFV Architecture Effects: Wild Card
2. Continued Global Data Center Virtualization
-- Public vs. Private Cloud

2016 Cisco and/or its affiliates. All rights reserved.

Cisco Global Cloud Index: Forecast and Methodology, 20152020

White Paper
Cisco Public

3. Cloud Service Trends

4. Workloads by Application
5. Data Center and Cloud Storage: Capacity and
Utilization
6. Global Digitization: Impact of IoE
-- Potential Effect of IoE on Global Data Centers
-- M2M Data Analytic Requirements Drive Fog/
Cloud Computing
7. Global Cloud Readiness
-- Security: Imperative for Cloud Growth
-- Network Speeds and Latency Analysis

Trend 1: Growth of Global Data Center

Relevance and Traffic
From server closets to large hyperscale deployments,
data centers are at the crux of delivering IT services
and providing storage, communications, and
networking to the growing number of networked
devices, users, and business processes in general.
The growing importance of data analyticsthe result
of big data coming from ubiquitously networked
end-user devices and IoE alikehas added to
the value and growth of data centers. They touch
nearly every aspect of an enterprise, whether
internal/employee-related data, communication
or processes, or partner- and customer-facing
information and services. The efficient and effective
use of data center technology such as virtualization,
new software-based architectures, and management
tools and use of public vs. private resources and so
on can all add to the agility, success, and competitive
differentiation of a business.
The increased focus on business agility and cost
optimization has led to the rise and growth of cloud
data centers. Cloud data centers have the five
essential characteristics of cloud computing as
listed by National Institute of Technology (NIST).
These five characteristics are on-demand selfservice, broad network access, resource pooling,
rapid elasticity or expansion, and measured
service. For more details, refer to Appendix E.
Cloud adoption enables faster delivery of services
and data, increased application performance, and
improved operational efficiencies.
Although security and integration with existing IT
environments continue to represent concerns for
some potential cloud-based applications, a growing
range of consumer and business cloud services are
currently available. Todays cloud services address
varying customer requirements (for example,
privacy, mobility, and multiple device access) and

2016 Cisco and/or its affiliates. All rights reserved.

support near-term opportunities as well as longterm strategic priorities for network operators, both
public and private.

Hyperscale Data Center Growth

The increasing need for data center and cloud

resources from both the business and consumer
service perspective has led to the development
of large-scale public cloud data centers called
hyperscale data centers. Hyperscale cloud operators
are increasingly dominating the cloud landscape.
To be a hyperscale cloud operator, a company must
meet the following criteria defined in terms of annual
revenues:
More than US$1 billion in annual revenue from
infrastructure as a service (IaaS), platform as a
service (PaaS), or infrastructure hosting services
(for example, Amazon/AWS, Rackspace, Google)
More than US$2 billion in annual revenue from
software as a service (SaaS) (for example,
Salesforce, ADP, Google)
More than US$4 billion in annual revenue from
Internet, search, and social networking (for
example, Facebook, Yahoo, Apple)
More than US$8 billion in annual revenue from
e-commerce/payment processing (for example,
Amazon, Alibaba, eBay)
Twenty-four hyperscale operators were identified
using the preceding criteria. The data centers
operated by these companies are what we consider
as hyperscale. The hyperscale operator might own
the data center facility, or it might lease it from a
colocation/wholesale data center provider.
Figure 1. Data Center Growth
47%

600

43%

33%

400
Hyperscale 300
Data Centers
200

45%

38%

500

27%

485

40%
35%

% Share
30% of Data
25% Center
Servers
20% (Installed
15% Base)

399
346

21%

259

447

50%

297

10%

100

5%
0%

0
2015

2016

2017

2018

2019

2020

Source: Cisco Global Cloud Index, 2015-2020; Synergy Research.

These hyperscale data centers will grow from 259

in number at the end of 2015 to 485 by 2020. They
will represent 47 percent of all installed data center
servers by 2020. In other words, they will account for
83 percent of the public cloud server installed base
in 2020 and 86 percent of public cloud workloads.

Cisco Global Cloud Index: Forecast and Methodology, 20152020

White Paper
Cisco Public

While only seven of these 24 companies are

headquartered outside of the United States, their data
center footprint is much more geographically diverse.
Figure 2. Data Center Growth: Regional View
600
Middle East and Africa
(0%,0.8%)
Central and Eastern
Europe (0.4%,1.4%)
Latin America (3.9%,
4.5%)
Western Europe (16%,
17%)
Asia Pacific (29%,
33%)
North America (51%,
43%)

500
400

Hyperscale 300
Data Centers
200
100
0
2015

2016

2017

2018

2019

2020

Note: Percentages within parentheses refer to relative share for 2015 and 2020.
Source: Cisco Global Cloud Index, 20152020; Synergy Research.

At the end of 2016, these 24 hyperscale operators

will in aggregate have 297 data centers, with North
America having the largest share, at 51 percent,
followed by Asia Pacific, with 29 percent, Western
Europe, with 17 percent, and Latin America, with 3
percent.
Asia Pacific has been the fastest growing region in
terms of hyperscale data center location and will
continue to grow more rapidly over the next five years,
although North America will still account for 43 percent
of hyperscale data centers by the end of 2020.
As with servers, hyperscale data centers represent
a large portion of overall data, traffic, and processing
power in data centers. Traffic within hyperscale data
centers will quintuple by 2020. Hyperscale data
centers already account for 34 percent of total traffic
within all data centers and will account for 53 percent
by 2020. Hyperscale data centers will also represent
57 percent of all data stored in data centers and 68
percent of total data center processing power.
Figure 3. The Scale of Hyperscale
By 2020, Hyperscale Data Centers Will House:

Today:

47%

of all data center servers

21%

68%

of all data center processing power

39%

57%

of all data stored in data centers

49%

53%

of all data center traffic

34%

Source: Cisco Global Cloud Index, 20152020.

Global Data Center IP Traffic: Three-Fold

Increase by 2020

Most Internet traffic has originated or terminated in

a data center since 2008, when peer-to-peer traffic
1

(which does not originate from a data center but

instead is transmitted directly from device to device)
ceased to dominate the Internet application mix.
Data center traffic will continue to dominate Internet
traffic for the foreseeable future, but the nature
of data center traffic is undergoing a fundamental
transformation brought about by cloud applications,
services, and infrastructure. The importance
and relevance of the global cloud evolution are
highlighted by one of the top-line projections
from this updated forecast: by 2020 more than 90
percent of data center traffic will be cloud traffic.
The following sections summarize not only the
volume and growth of traffic entering and exiting
the data center, but also the traffic carried between
different functional units within the data center,
cloud versus traditional data center segments, and
business versus consumer cloud segments.
Figure 4 summarizes the forecast for data center IP
traffic growth from 2015 to 2020.
Figure 4. Global Data Center IP Traffic Growth
27% CAGR
2015-2020
18
16
14
12
Zettabytes 10
per Year 8
6
4
2
0

15.3
12.9
10.8
8.6
6.5
4.7

2015

2016

2017

2018

2019

2020

Source: Cisco Global Cloud Index, 20152020.

Although the amount of global traffic crossing

the Internet and IP WAN networks is projected to
reach 2.3 ZB per year by 20201, the amount of
annual global data center traffic in 2015 is already
estimated to be 4.7 ZB and by 2020 will triple to
reach 15.3 ZB per year. This increase represents a
27 percent CAGR. The higher volume of data center
traffic is due to the inclusion of traffic inside the data
center (typically, definitions of Internet and WAN
traffic stop at the boundary of the data center).
The global data center traffic forecast, a major
component of the Cisco GCI report, covers
network data centers worldwide operated by
service providers as well as enterprises. Refer to
Appendix A for more details about the methodology
of the data center and cloud traffic forecasts and
Appendix B for the positioning of the GCI Forecast
relative to the Cisco VNI Global IP Traffic Forecast.

Refer to Cisco Visual Networking Index: Forecast and Methodology, 20152020.

2016 Cisco and/or its affiliates. All rights reserved.

Cisco Global Cloud Index: Forecast and Methodology, 20152020

White Paper
Cisco Public

Table 1 provides details for global data center traffic growth rates.
Table 1.

Global Data Center Traffic, 20152020

Data Center IP Traffic, 20152020

2015

2016

2017

2018

2019

2020

CAGR
20152020

Data center to user

744

933

1,164

1,438

1,772

2,183

24.0%

Data center to data center

346

515

713

924

1,141

1,381

31.9%

Within data center

3,587

5,074

6,728

8,391

10,016

11,770

26.8%

Consumer

2,997

4,304

5,836

7,435

9,075

10,906

29.5%

Business

1,681

2,218

2,768

3,318

3,853

4,429

21.4%

Cloud data center

3,851

5,636

7,712

9,802

11,850

14,076

29.6%

Traditional data center

827

885

892

951

1,078

1,259

8.8%

4,678

6,522

8,604

10,753

12,928

15,335

26.8%

By Type (EB per Year)

By Segment (EB per Year)

By Type (EB per Year)

Total (EB per Year)

Total data center traffic
Definitions:
Data center to user: Traffic that flows from the
data center to end users through the Internet or
IP WAN
Data center to data center: Traffic that flows
from data center to data center
Within data center: Traffic that remains within the
data center, excludes traffic within the rack
Consumer: Traffic originating with or destined for
consumer end users
Business: Traffic originating with or destined for
business end users
Cloud data center: Traffic associated with cloud
data centers
Traditional data center: Traffic associated with
noncloud data centers

2016 Cisco and/or its affiliates. All rights reserved.

Data Center Traffic Destinations: Most Traffic

Remains Within the Data Center

Consumer and business traffic flowing through data

centers can be broadly categorized into three main
areas (Figure 5):
Traffic that remains within the data center:
For example, moving data from a development
environment to a production environment within a
data center, or writing data to a storage array
Traffic that flows from data center to data
center: For example, moving data between
clouds, or copying content to multiple data
centers as part of a content distribution network
Traffic that flows from the data center to
end users through the Internet or IP WAN:
For example, streaming video to a mobile device
or PC

Cisco Global Cloud Index: Forecast and Methodology, 20152020

White Paper
Cisco Public

Figure 5. Global Data Center Traffic by Destination in 2020

Data Center
to user
14%

Within Data Center

77%

Data Center
to
Data Center
9%

Within Data Center (77%)

2020
Storage, production and
development data,
authentication

Data Center to Data Center (9%)

Replication, CDN,
intercloud links

Data Center to User (14%)

Web, email
internal VoD,
WebEx...

2020
Source: Cisco Global Cloud Index, 20152020.

The portion of traffic residing within the data center

will remain at the same level over the forecast
period, accounting for 77 percent of data center
traffic in 2015 and about 77 percent by 2020. The
totals for within the data center do not include racklocal traffic (traffic that remains within a given server
rack). Rack-local traffic is approximately twice the
size of the within data center volumes shown in
the forecast. The inclusion of rack-local traffic would
change our traffic distribution to show more than 90
percent of traffic remaining local to the data center.
Big data is a significant driver of traffic within the
data center. While much of big data traffic is racklocal, enough exits the rack that big data will be
responsible for 17 percent of all traffic within the data
center by 2020, up from 10 percent in 2015. Video
does not drive a large volume of traffic within the
data center, since minimal processing is done on the
video relative to the large size of the video stream.
Traffic between data centers is growing faster than
either traffic to end users or traffic within the data
center, and by 2020, traffic between data centers
will account for almost 9 percent of total data center
traffic, up from 7 percent at the end of 2015. The
high growth of this segment is due to the increasing
prevalence of content distribution networks, the
proliferation of cloud services and the need to shuttle
data between clouds, and the growing volume of
data that needs to be replicated across data centers.
Overall, east-west traffic (traffic within the data
center and traffic between data centers) represents
86 percent of total data center by 2020,

2016 Cisco and/or its affiliates. All rights reserved.

and north-south traffic (traffic exiting the data

center to the Internet or WAN) is only 14 percent of
traffic associated with data centers.

Global Data Center and Cloud IP Traffic Growth

Data center traffic on a global scale will grow at a
27 percent CAGR (Figure 4), but cloud data center
traffic will grow at a faster rate (30 percent CAGR)
or 3.7-fold growth from 2015 to 2020 (Figure 6).
Figure 6. Total Data Center Traffic Growth
18
16
14
12
10
Zettabytes
8
per Year
6
4
2
0

27% CAGR
2015-2020

Traditional Data Center (9% CAGR)

Cloud Data Center (30% CAGR)
8%

92%
18%
82%

2015

2016

2017

2018

2019

2020

Source: Cisco Global Cloud Index, 20152020.

Figure 7. Cloud Data Center Traffic Growth

16
14
12
10
Zettabytes 8
per Year 6
4
2
0

30% CAGR
2015-2020
14.1
11.9
9.8
7.7
5.6
3.9

2015

2016

2017

2018

2019

2020

Source: Cisco Global Cloud Index, 20152020.

Cisco Global Cloud Index: Forecast and Methodology, 20152020

White Paper
Cisco Public

Cloud will represent more than 90 percent of all data

center traffic will be based in the cloud. (For regional
cloud traffic trends, refer to Appendix C) Significant
promoters of cloud traffic growth include the rapid
adoption of and migration to cloud architectures
and the ability of cloud data centers to handle
significantly higher traffic loads. Cloud data centers
support increased virtualization, standardization, and
automation. These factors lead to better performance
as well as higher capacity and throughput.

The Evolution of Data Center Architecture:

SDN/NFV

Three technology trends are transforming the data

center: leaf-spine architectures (which flatten the
tiered architecture of the data center), softwaredefined networks (SDNs, which separate the control
and forwarding of data center traffic), and network
function virtualization (NFV, which virtualizes a
variety of network elements).
Most major hyperscale data centers already employ
flat architectures and software-defined network
and storage management, and adoption of SDN/
NFV within large-scale enterprise data centers has
been rapid. As a portion of traffic within the data
center, SDN/NFV is already transporting 23 percent,
growing to 44 percent by 2020 (Figure 8).
Figure 8. SDN/NFV Traffic Within the Data Center
6.0

41%
37%

5.0

5.2

33%

Zettabytes
per Year

4.0
3.0

44%

23%

30%

% Share of
Within
Data
20%
Center
15%
Traffic
25%

3.1

2.0

2.2

10%

1.4

1.0

40%
35%

4.1

28%

50% 44% CAGR

20152020
45%

5%

0.8

0%

0.0
2015

2016

2017

2018

2019

2020

Big data: Traffic engineering enabled by SDN/

NFV supports elephant2 data flows without
compromising mouse3 data flows, making it
safe to transport large amounts of data to and
from big data clusters.
Video bitrates: SDN will allow video bitrates to
increase, because SDN can seek out highest
bandwidth available even midstream, instead
of lowering the bitrate according the available
bandwidth for the duration of the video, as is
done today.
Cloud gaming: SDN can decrease latency within
the data center, decreasing delay experiences by
end-users in cloud gaming applications, which
might help increase cloud gaming adoption by
both content providers and end users.

Trend 2: Continued Global Data Center

Virtualization
A server workload is defined as a virtual or physical
set of computer resources, including storage,
that are assigned to run a specific application or
provide computing services for one to many users.
A workload is a general measurement used to
describe many different applications, from a small
lightweight SaaS application to a large computational
private cloud database application. For the purposes
of quantification, we consider each workload being
equal to a virtual machine or a container. In fact,
containers are one of the factors enabling a steady
increase in the number of workloads per server
deployed. The Cisco Global Cloud Index forecasts
the continued transition of workloads from traditional
data centers to cloud data centers. By 2020, 92
percent of all workloads will be processed in cloud
data centers (Figure 9). For regional distributions of
workloads, refer to Appendix D.
Figure 9. Workload Distribution: 20152020

Source: Cisco Global Cloud Index, 20152020.

SDN and NFV, along with flat architectures, might

streamline traffic flows with the data center such
that traffic is routed more efficiently in the future
than it is routed today. In theory, SDN allows for
traffic handling policies to follow virtual machines
and containers, so that those elements can be
moved within a data center in order to minimize
traffic in response to bandwidth bottlenecks.
However, there are also ways in which SDN/NFV
can lead to an increase in both data center traffic
and in general Internet traffic:

600

Traditional Data Center (3% CAGR)

21% CAGR
2015-2020

Cloud Data Center (26% CAGR)

8%

500

92%

400
Installed
Workload 300
in Millions

200
100
0

25%
75%

2015

2016

2017

2018

2019

2020

Source: Cisco Global Cloud Index, 20152020.

Elephant flows have varying definitions in the industry, but refer to flows of traffic that carry a disproportionate amount
of traffic in terms of bytes, usually greater than 1% of total traffic in a time period.

Mouse flows generate average-or-below traffic, but might have strict requirements in terms of latency.

2016 Cisco and/or its affiliates. All rights reserved.

Cisco Global Cloud Index: Forecast and Methodology, 20152020

White Paper
Cisco Public

Cloud workloads are expected to more than triple (grow 3.2-fold) from 2015 to 2020, whereas traditional
data center workloads are expected to see a global decline, at a negative 3 percent CAGR from 2015 to
2020. Traditionally, one server carried one workload. However, with increasing server computing capacity
and virtualization, multiple workloads per physical server are common in cloud architectures. Cloud
economics, including server cost, resiliency, scalability, and product lifespan, along with enhancements
in cloud security, are promoting migration of workloads across servers, both inside the data center and
across data centers (even data centers in different geographic areas). Often an end-user application can
be supported by several workloads distributed across servers. Table 2 provides details about the shift of
workloads from traditional data centers to cloud data centers.
Table 2.

Workload Shift from Traditional Data Centers to Cloud Data Centers

Global Data Center Workloads in Millions

2015

2016

2017

2018

2019

2020

CAGR
20152020

Traditional data center workloads

44.9

45.1

44.2

43.4

41.5

38.8

-3%

Cloud data center workloads

136.0

189.8

255.4

322.0

383.3

440.0

26%

Total data center workloads

180.9

234.9

299.7

365.4

424.8

478.8

21%

Cloud workloads as a percentage

of total data center workloads

75%

81%

85%

88%

90%

92%

Traditional workloads as a
percentage of total data center
workloads

25%

19%

15%

12%

10%

8%

One of the main factors prompting the migration

of workloads from traditional data centers to cloud
data centers is the greater degree of virtualization
(Figure 10) in the cloud space, which allows
dynamic deployment of workloads in the cloud to
meet the dynamic demand of cloud services. This
greater degree of virtualization in cloud data centers
can be expressed as workload density. Workload
density measures average number of workloads
per physical server. The workload density for cloud
servers will grow from 7.3 in 2015 to 11.9 by 2020.
In comparison, the workload density in traditional
data center servers will grow from 2.2 in 2015 to
3.5 by 2020.
Figure 10. Increasing Cloud Virtualization
14

Traditional Data Center

12

Cloud Data Center

11.9

10
Average
Workload
Density

7.3

3.4X
3.3X

4
2

Workloads
per Server

3.5

2.2

2015

2016

2017

2018

2019

Workloads
per Server

2020

Source: Cisco Global Cloud Index, 20152020.

Public vs. Private Cloud4

We look into the growth of public cloud vs. private

cloud through workload analysis. Public cloud, as
indicated by the workloads growth, is growing faster
than the private cloud. As the business sensitivity
to costs associated with dedicated IT resources
grows along with demand for agility, we can see a
greater adoption of public cloud by the businesses,
especially with strengthening of public cloud
security. Although many mission-critical workloads
might continue to be retained in the traditional data
centers or private cloud, the public cloud adoption
is increasing along with the gain in trust in public
cloud. Some enterprises might adopt a hybrid
approach to cloud. In a hybrid cloud environment,
some of the cloud computing resources are
managed in-house by an enterprise and some are
provided by an external provider. Cloud bursting is
an example of hybrid cloud where daily computing
requirements are handled by a private cloud, but
for sudden spurts of demand the additional traffic
demand (bursting) is handled by a public cloud.
While the overall cloud workloads are growing
at a CAGR of 26 percent from 2015 to 2020
(Figure 11), the public cloud workloads are going to
grow at 35 percent CAGR from 2015 to 2020, and

For definition of public and private cloud, refer to Appendix E.

2016 Cisco and/or its affiliates. All rights reserved.

Cisco Global Cloud Index: Forecast and Methodology, 20152020

White Paper
Cisco Public

private cloud workloads will grow at a slower

pace of 15 percent CAGR from 2015 to 2020.
By 2016 there will be more workloads (56 percent)
in the public cloud as compared to private cloud
(44 percent).
Figure 11. Public vs. Private Cloud Growth
500
450
400
350
300
Installed 250
Workloads 200
in Millions
150
100
50
0

Public Cloud Data Center (35% CAGR)

26% CAGR
2015-2020

Private Cloud Data Center (15% CAGR)

68%

49%
51%

56%
32%

44%

2015 2016* 2017

2018

2019

2020

Source: Cisco Global Cloud Index, 20152020.

This growth of workloads in the public cloud space

is also reflected in the growth of virtualization, as
shown in Figure 12. The workload density in public
cloud data centers will overtake that in private cloud
data centers by 2016.
Figure 12. Public Cloud Virtualization Gaining Momentum
14

Traditional Data Center

12

Public Cloud Data Center

10
Average
Workload
Density

12.4

Private Cloud Data Center

11.0

7.7
6.9

Figure 13. SaaS Most Highly Deployed Global Cloud Service

from 2015 to 2020
500
450
400
350
300
Installed 250
Workloads
in Millions 200
150
100
50
0

Saas (30% CAGR)

26% CAGR
2015-2020

IaaS (17% CAGR)

PaaS (24% CAGR)

8%
17%

74%

9%
26%
65%

2015

2016

2017

2018

2019

2020

3.5
2.2

2015

2016

2017

2018

2019

2020

Source: Cisco Global Cloud Index, 20152020.

Trend 3: Cloud Service Trends

This section reviews the growth of the three
different cloud service categories: IaaS, PaaS, and
SaaS5. Although numerous other service categories
have emerged over time, they can be aligned
within the IaaS, PaaS, and SaaS categorization. For
example, business process as a service (BPaaS) is
considered part of SaaS. For simplicity we can think
of these three service models as different layers of
cloud with infrastructure at the bottom, the platform
next, and finally software at the top.

At the aggregate cloud level, we find that SaaS

workloads maintain majority share throughout the
forecast years, and by 2020 will have 74 percent
share of all cloud workloads, growing at 30 percent
CAGR from 2015 to 2020 (Figure 13). PaaS will
have the second-fastest growth, although it will lose
the share of total cloud workloads from 9 percent in
2015 to 8 percent by 2020.

Source: Cisco Global Cloud Index, 20152020.

4
2

GCI categorizes a cloud workload as IaaS, PaaS, or

SaaS based upon how the user ultimately uses the
service, regardless of other cloud services types that
might be involved in the final delivery of the service.
As an example, if a cloud service is a SaaS type
but it also depends on some aspects of other cloud
services such as PaaS or IaaS, such a workload is
counted as SaaS only. As another example, if a PaaS
workload operates on top of an IaaS service, such a
workload is counted as PaaS only.

In order to understand the reasons behind this trend,

we have to analyze the public and private cloud
segments a bit more deeply. In the private cloud,
initial deployments were predominantly IaaS. Test and
development types of cloud services were the first to
be used in the enterprise; cloud was a radical change
in deploying IT services, and this use was a safe and
practical initial use of private cloud for enterprises. It
was limited, and it did not pose a risk of disrupting the
workings of IT resources in the enterprise. As trust
in adoption of SaaS or mission-critical applications
builds over time with technology enablement in
processing power, storage advancements, memory
advancements, and networking advancements, we
foresee the adoption of SaaS type applications to
accelerate over the forecast period (Figure 14), while
shares of IaaS and PaaS workloads decline.

For definition of IaaS, PaaS, and SaaS, refer to Appendix E

2016 Cisco and/or its affiliates. All rights reserved.

10

Cisco Global Cloud Index: Forecast and Methodology, 20152020

White Paper
Cisco Public

Trend 4: Workloads by Application

Figure 14. SaaS Gains Momentum in Private Cloud

160

Paas (2% CAGR)

140

IaaS (-6% CAGR)

120

15% CAGR
2015-2020

SaaS (27% CAGR)

7%
14%

100
Installed
Workloads
in Millions

80
60

13%

40

39%

20

48%

2015

79%

2016

2017

2018

2019

This is the first year we are looking at workloads

split by applications. We estimate that in 2015
enterprises (including SMB, government, and public
sector) accounted for 79 percent of workloads and
consumers 21 percent. Consumer share of the total
will grow to 28 percent by 2020, while enterprise
sector share will decline to 72 percent (Figure 16).
Figure 16. Global Data Center Workloads: Consumer vs.
Enterprise Applications

2020

Source: Cisco Global Cloud Index, 20152020.

Consumer Workloads (28% CAGR)

In the public cloud segment the first cloud services

to be deployed were SaaS. SaaS services offered
in the public cloud were often a low-risk and easyto-adopt proposition, with some clear financial
and flexibility benefits to users. The first users of
SaaS were the consumer segment, followed by
some small and medium-sized businesses. As
public SaaS solutions become more sophisticated
and robust, larger enterprises are adopting these
services as well, beginning with less-critical
services. Enterprises, especially large ones, will
be carefully weighing the benefits (scalability,
consistency, cost optimization, and so on) of
adopting public cloud services against the risks
(security, data integrity, business continuity, and so
on) of adopting such services.
As shown in Figure 15, IaaS and PaaS have gone
beyond the initial stages of deployment in the public
cloud. Spend on public IaaS and PaaS is still small
compared with spend on enterprise data center
equipment, data center facilities, and associated
operating expenses. These cloud services will
gain momentum over the forecast period as more
competitive offers come to the market and continue
to build enterprise trust for outsourcing these more
technical and fundamental services.
Figure 15. IaaS and PaaS Gain Public Cloud Share of
Workloads
35% CAGR
2015-2020

350

Paas (51% CAGR)

300

IaaS (48% CAGR)

9%

SaaS (31% CAGR)

19%

250
Installed 200
Workloads
in Millions 150

100
50
0

72%
5%
12%
83%

2015

2016

2017

2018

2019

2020

600
500

28%

Installed 400
Workloads
300
in Millions

200
100
0

21% CAGR
2015-2020

Enterprise Workloads (19% CAGR)

72%

21%
79%

2015

2016

2017

2018

2019

2020

Source: Cisco Global Cloud Index, 20152020.

Within enterprise, compute/IaaS and collaboration

are the two main contributors to workload totals,
while on the consumer side social networking and
video/media streaming are the biggest contributors.
While the percentage mix will change, those will
remain the biggest contributors to workload totals
over the next five years (Figure 17). For definitions
of the applications, see Appendix F.
Figure 17. Global Data Center Workloads by Applications

600
500
400
Installed
Workloads
300
in Millions

Other Consumer Apps (23% CAGR)

Search (25% CAGR)
Social Networking (33% CAGR)
Video Streaming (32% CAGR)
ERP and Other Business Apps (17% CAGR)
Database/ Analytics / IoT (22% CAGR)
Collaboration (18% CAGR)
Compute (21% CAGR)

21% CAGR
2015-2020

Consumer

200

Enterprise

100
0

2015

2016

2017

2018

2019

2020

Source: Cisco Global Cloud Index, 20152020.

If we look at the application split of workloads across

traditional, public, and private cloud data centers,
then we find that public cloud data centers have the
largest share of consumer application workloads,
while traditional and private cloud data centers have
a larger share in the business/enterprise segment
(Figure 18).

Source: Cisco Global Cloud Index, 20152020.

2016 Cisco and/or its affiliates. All rights reserved.

11

Cisco Global Cloud Index: Forecast and Methodology, 20152020

White Paper
Cisco Public

Figure 18. Global Data Center Workloads by Applications:

Traditional vs. Cloud (2015)
Traditional

Private Cloud

Figure 20. Global Data Center Storage Capacity:

Traditional vs. Cloud
2,000

Public Cloud

Other Consumer Apps

1,800

Search

1,600

1,842
1,405

1,400

Social Networking

1,200
Installed
Storage 1,000
Capacity in 800
Exabytes 600

Video Streaming
ERP and Other Business Apps
Database/ Analytics / IoT
Collaboration

400

Compute

200

0%

Public Cloud Data Center

Private Cloud Data Center
Traditional Data Center

20%

40%

60%

80%

100%

70%

1,065
782
382
45%
20%
35%

Source: Cisco Global Cloud Index, 20152020.

2015

546
18%
12%
2016

2017

2018

2019

2020

Source: Cisco Global Cloud Index, 20152020.

However, by 2020, traditional and private cloud data

centers will lose share to public clouds across all
applications except for database/analytics/IoE and
other business application categories. In the latter
categories, private clouds will mostly maintain their
share (Figure 19).
Figure 19. Global Data Center Workloads by Applications:
Traditional vs. Cloud (2020)
Traditional

Private Cloud

Public Cloud

Global Data Center Storage Utilization

Also for the first time we have estimated the total

amount of actual data stored within data centers.
Globally, the data stored in data centers will
quintuple by 2020 to reach 915 EB by 2020, up
5.3-fold (a CAGR of 40%) from 171 EB in 2015
(Figure 21).
Figure 21. Actual Data Stored in Data Centers

Other Consumer Apps

1,000

Search

900

Social Networking

800

Video Streaming

700

40% CAGR
2015-2020
915

600

ERP and Other Business Apps

Exabytes

Database/Analytics/IoT

689

500

513

400

370

300

Collaboration

200

Compute
0%

20%

40%

60%

80%

100%

Source: Cisco Global Cloud Index, 20152020.

Trend 5: Data Center and Cloud Storage6:

Capacity and Utilization
Global Data Center Storage Installed Capacity
This year, for the first time, we have looked at the
installed storage capacity in global data centers.
We estimate that total data center storage capacity
will grow nearly 5-fold from 2015 to 2020, growing
from 382 EB in 2015 to 1.8 ZB by 2020. Cloud will
account for 88 percent of the total storage capacity
(Figure 20).

100
0

251
171

2015

2016

2017

2018

2020

Big data is a key driver of overall growth in stored

data. Big data will reach 247 EB by 2020, up almost
10-fold from 25 EB in 2015. Big data alone will
represent 27 percent of data stored in data centers
by 2020, up from 15 percent in 2015 (Figure 22).
Figure 22. Big Data Volumes
58% CAGR
2015-2020

300
250
247

200

Exabytes 150

173

100

116
73

50
0

2019

43

2015

2016

2017

2018

2019

2020

Storage does not include archival media such as tape.

2016 Cisco and/or its affiliates. All rights reserved.

12

Cisco Global Cloud Index: Forecast and Methodology, 20152020

White Paper
Cisco Public

Big data is defined here as data deployed in a

distributed processing and storage environment
(such as Hadoop). Generally speaking, distributed
processing is chosen as a data architecture when
the data is big in volume (more than 100 terabytes),
velocity (coming in or going out at more than
10 gigabytes per second), or variety (combining
data from a dozen or more sources). Big data is
sometimes used interchangeably with data analytics
or data science, but data science techniques can be
used on data of any size, and the quality of insights
achieved with data science is not related to the size
of the underlying data.
As large as the data stored in data centers will be
(nearly 1 zettabyte by 2020), the amount of data
stored on devices will be 5 times higher: 5.3 ZB by
2020. Out of the combined 6.2 ZB of stored data in
the world, most stored data will continue to reside
in client devices, as it does today. Today, only 12
percent of total stored data is stored in the data
center, but more data will move to the data center
over time (Figure 23). In addition to larger volumes
of stored data, the stored data will be coming from
a wider range of devices by 2020. Currently, 61
percent of data stored on client devices resides
on PCs. By 2020, stored data on PCs will reduce
to 52 percent, with a greater portion of data on
smartphones, tablets, and machine-to-machine
(M2M) modules. Stored data associated with
M2M grows at a faster rate than any other device
category.
Figure 23. Data Center Storage Analysis
The volume of all data stored will almost triple by 2020 from 1.4 ZB to 6.2 ZB.
Most data is stored on client devices, but more moves to the data center over time.
2015

12% DC

88% Client Devices or M2M

2020

1.4 ZB

84% Client Devices or M2M

Data Stored on Client Devices (PCs, Tablets, Phones, M2M...)

2015
2020

16% DC

Data Stored on Data Centers

61%
52%
PCs
Tablets

18%

6.2 ZB

12% 8% 5%5%
Smartphones + Phablets
M2M

915
EB

Consumer Cloud Storage Growth

Along with the growth in consumer Internet

population and multidevice ownership devices,
we are seeing a significant growth in the use of
cloud services such as consumer cloud storage,
also called personal content lockers. In personal
content lockers, users can store and share music,
photos, and videos through an easy-to-use
interface at relatively low or no cost. Furthermore,
the proliferation of tablets, smartphones, and other
mobile devices allows access to personal content
lockers in a manner convenient to the user.
Cisco GCI estimates that by 2020, 59 percent (2.3
billion) of the consumer Internet population will use
personal cloud storage, up from 47 percent (1.3
billion users) in 2015 (Figure 24).
Figure 24. Personal Cloud Storage: Growth in Users

2,500
2,000

1,000

1,329

1,561

2,309

2,111

1,754

500
0
2015

2016

2017

2018

2019

2020

Source: Cisco Global Cloud Index, 20152020; Juniper Research.

Cisco GCI forecasts that global consumer cloud

storage traffic will grow from 8 EB annually in 2015
to 48 EB by 2020 at a CAGR of 42 percent (Figure
25). This growth translates to per-user traffic of 1.7
gigabytes (GB) per month by 2020, compared to
513 MB per month in 2015.
Figure 25. Consumer Cloud Storage Traffic7 Growth
60

2020
Data at Rest (Stored)
Data in Motion (Traffic)

42% CAGR
2015-2020

50
40

Source: Cisco Global Cloud Index, 20152020.

1,926

1,500

Consumers
in Millions

981
EB
per
mo.

Over time, cloud-based services will enable

consumers and businesses alike to move more
of their stored data to a central repository that
can provide ubiquitous access to content and
applications through any device at any location. The
following section covers the growth in consumer
cloud storage.

12% CAGR
2015-2020

48
37

30
Exabytes
per Year

29

20

21
14

10
0

8
2015

2016

2017

2018

2019

2020

Source: Cisco Global Cloud Index, 20152020.

Consumer cloud storage traffic includes personal content lockers, cloud backup, and so on, a
 nd does not include
cloud DVR.

2016 Cisco and/or its affiliates. All rights reserved.

13

Cisco Global Cloud Index: Forecast and Methodology, 20152020

White Paper
Cisco Public

Trend 6: Global Digitization: Impact of IoE

Potential Effects of IoE on Global Data Centers
Cloud services are accelerated in part by the unprecedented amounts of data being generated by not only
people but also machines and things. Cisco GCI estimates that 600 ZB will be generated by all people,
machines, and things by 2020, up from 145 ZB generated in 2015.
Figure 26 shows examples of the amounts of data that will be generated by planes, automobiles, and
buildings, among other things and systems.
Figure 26. Smart City: Multiple Applications Create Big Data

What Makes a Smart City?

Multiple Applications Create Big Data

Connected Plane

Intelligent Building

40 TB per day (0.1% transmtted)

275 GB per day (1% transmitted)

Connected Factory
1 PB per day (0.2% transmitted)

Public Safety
50 PB per day (<0.1% transmitted)

A city of
one million
will generate
200 million gigabytes
of data per day
by 2020

Smart Hospital
5 TB per day (0.1% transmitted)

Smart Car
70 GB per day (0.1% transmitted)

Weather Sensors

Smart Grid

10 MB per day (5% transmitted)

5 GB per day (1% transmitted)

Source: Cisco Global Cloud Index, 20152020

Most of the more than 600 ZB that will be generated by 2020 will be ephemeral in nature and will be neither
saved nor stored. Much of this ephemeral data is not useful to save, but we estimate that approximately
10 percent is useful, which means that there will be 10 times more useful data being created (60 ZB, 10
percent of the 600 total) than will be used (6 ZB) in 2020. Edge or fog computing might help bridge this gap.

M2M Data Analytic Requirements Drive Fog/Cloud Computing

Growth in machine-to-machine connections and applications is also driving new data analytics needs.
Although not all M2M applications promote a lot of traffic, the sheer number of these connections is capable of
delivering intelligent, actionable information if the data from them can be analyzed. Figure 27 maps several of
such M2M applications for their complexity of protocol environment and data analytic needs. IoT applications
have very different characteristics. In some cases application analytics and management can occur at the edge
device level whereas for others it is more appropriately handled centrally, typically hosted in the cloud.
The opportunity for fog propositions (distributed intelligence) and the associated intelligent gateways
is the strongest in markets that meet two conditions. First, their focus of data analytics is on the socalled aggregation level, and, second, they are subject to a problematic degree of protocol complexity.
Especially the first aspect is set to be decisive in determining the demand for intelligent gateways as a
product segment. In general, such markets reside in the Industrial IoT domain involving verticals such as
manufacturing, extractive industries, and healthcare. Applications such as smart metering can benefit from
real-time analytics of aggregated data that can optimize the usage of resources such as electricity, gas, and
water. Local level analytics is suited for those applications that require the data to be stored and analyzed
locally due to either regulatory reasons or because the cost of transportation of the data upstream and
associated wait-time for analysis is prohibitive.
A key issue for IoT in the coming years is subsidiarity, i.e. performing the data analysis at the appropriate
level. In most cases, there will be a blend of approaches and the functionality to manage local as well as
central application management will be increasingly critical.
2016 Cisco and/or its affiliates. All rights reserved.

14

Cisco Global Cloud Index: Forecast and Methodology, 20152020

White Paper
Cisco Public

Aggregation
Level

Usage Based
Insurance

Water
Metering

Home
Automation

Electricity
Metering

Parking
Space Mgmt

Smart Grid
(Water)
Warehousing
and Storage

Connected Car
(Autonomous
Driving)

Venue
Management

CCTV
(Event Detection)

Mining
Operation

Transport
(Air)

Consolidated

Transport
(Road)

Smart Grid
(Electricity)

Street Lighting

Local
Level

Focus of
Data Analytics

Cloud
Level

Figure 27. Growth in M2M Connections Drive New Data Analytics Needs

Transport
(Rail)

Oil and Gas

(Upstream)
Transport
(Sea)

Manageable

Building
Automation
Manufacturing
(Discrete)
Manufacturing
(Process)
Healthcare
(Equipment)

Problematic

Complexity of
Protocol Environment

Source: Cisco Global Cloud Index, 20152020

Trend 7: Global Cloud Readiness

Security: Imperative for Cloud Growth

The move to the cloud is imminent. In just the past year, a variety of businesses and organizations have
reported their plans for cloud migration or adoption. For example, Netflix Inc. announced plans to shut down
the last of its traditional data centers during 2016, a step that will make it one of the first big companies to
run all of its IT in the public cloud. For our streaming business, we have been 100 percent cloud-based for
customer facing systems for some time now, and are planning to completely retire our data centers later this
summer.8 Several additional cloud examples are provided in Figure 28.
Figure 28. Examples of Broad Cloud Adoption

Examples of Broad Cloud Adoption

Barriers overcome & operational efficiency prevails

Netflix

Netflix Closes last DC,

Completes Cloud Migration

34.3% of Health Information Exchange

is Already in the Cloud

We rely on the cloud for all of our scalable computing and storage
needs our business logic, distributed databases and big data
processing/analytics, recommendations, transcoding, and hundreds of
other functions.
Netflix Representative

Healthcare organizations are increasingly willing to trust the cloud with

Protected Health Information (PHI). 36.2% of patient engagement tools
are in the cloud and 5.3% of the organizations leverage cloud for
compute cycles to analyze big data.
2016 HIMSS Analytics Cloud Survey

Deutshce Banks to move 30% of

Banke Workloads to cloud in 3 years

HESS
GE Oil & Gas

Pressure to cut infrastructure costs, increased flexibility, paired with

security and compliance services from the cloud vendors has boosted
banks willingness to explore the technology
Wall Street Journal

The integrated oil company is undergoing a large scale migration to

public cloud. Pretty much anything and everything were setting up,
they have sensors.
Zhanna Golodryga, Hess CIO

More than 50% of Workloads

in the Cloud

GE oil and gas migrated half of its core application to the cloud.

Scalability and allocation of resources are the major advantages of virtualization (refer to the section Trend
2: Continued Global Data Center Virtualization) and cloud computing. Administrators can bring up virtual
machines (VMs) and servers quickly without having the overhead of ordering or provisioning new hardware.
Hardware resources can be reassigned quickly and extra processing power can be consumed by other
services for maximum efficiency. By taking advantage of all the available processing power and untethering
the hardware from a single server model, cost efficiencies are being realized in both private and public clouds.
8

http://blogs.wsj.com/cio/2016/08/14/the-morning-download-netflix-leads-way-into-cloud-closing-final-data-center/

2016 Cisco and/or its affiliates. All rights reserved.

15

Cisco Global Cloud Index: Forecast and Methodology, 20152020

White Paper
Cisco Public

According to the National Institute of Technology

(NIST), cloud computing can be divided into three
main service types (refer to the section Trend 3:
Cloud Service Trends): Infrastructure as a Service
(IaaS), Platform as a Service (PaaS), and Software
as a Service (SaaS), and each affects data control
and governance a little differently. With IaaS, the
customer might have full control of the actual server
configuration, granting them more risk management
control over the environment and data. In PaaS,
the provider manages the hardware and underlying
operating system, limiting enterprise risk management
capabilities on those components. With SaaS, both the
platform and the infrastructure are fully managed by
the cloud provider, meaning if the underlying operating
system or service isnt configured appropriately the
data in the higher layer application might be at risk.
Cybercrime damages will cost the world $6 trillion
annually by 2021, up from $3 trillion in 2015. The
cybercrime costs prediction includes damage and
destruction of data, stolen money, lost productivity,
theft of intellectual property, theft of personal and
financial data, embezzlement, fraud, postattack
disruption to the normal course of business, forensic
investigation, restoration and deletion of hacked data
and systems, and reputational harm.9 Cyberthreats
have evolved from targeting and harming computers,
networks, and smartphones to people, cars, railways,
planes, power grids, and anything with a heartbeat
or an electronic pulse, all powered by the cloud.
The last several years have undoubtedly been the
most eventful period from a cloud security threat
perspective, with various instances of massive
breaches and escalating distributed denial-of-service
(DDoS) amplification attacks. Many network security
vendors and other networking device manufacturers
raced to patch their appliances against Heartbleed,
a serious vulnerability in the popular OpenSSL
cryptographic software library, and Shellshock, an
open-source vulnerability, which set off a series of
patch releases, among many others. The response
showcased the effectiveness of security vendors
to provide support and assistance to customers in
need. It also shed light on the broad adoption and
very quickly evolving landscape of cloud security
and management. Across all documents uploaded
to file sharing services, the most common type
of sensitive content is confidential company data
(for example, financial records, business plans,
source code, trading algorithms, and so on). A
total of 7.6 percent of documents in file sharing
services contain confidential data. Thats followed
by personally identifiable information (for example,
Social Security numbers, tax ID numbers, phone
9

numbers, addresses, and so on) at 4.3 percent of all

documents. Next, 2.3 percent of documents contain
payment data (for example, credit card numbers,
debit card numbers, bank account numbers, and
so on). Finally, 1.6 percent of documents contain
protected health information (for example, patient
diagnoses, medical treatments, medical record IDs,
and so on).10
Users expect their online experiences to be always
available and always secureand their personal and
business assets to be safe. As more data, business
processes, and services move to the cloud,
organizations are challenged to protect websites
and infrastructure without sacrificing performance
for security.
To help meet user expectations, more secure
Internet servers are being deployed worldwide. This
situation creates a growing infrastructure footprint
that provides more stringent authorization and
authentication processes and better serves end
users with secure transactions and communication.
The percentage of secure Internet servers that
conduct encrypted transactions over the Internet
using a Secure Sockets Layer (SSL) to the total
number of web facing servers is shown in Figure 29.
In the past year, North America and Western Europe
led with the percentage of secure Internet servers
compared to web-facing Internet servers.
Figure 29. Percentage of Secure Internet Servers to Total
Web-Facing Internet Servers by Region and Increase from
End of Year 2014 to 2015
Enabling Authentication and Secure Internet

Percentage of secure internet servers to all web facing servers

West Europe
51%
+1% YoY

North America
28%
+1% YoY

Asia Pacific
24%
+1% YoY

MEA
12%
+2% YoY

Central and
Eastern Europe
34%
+5% YoY

Latin
America
14%
+1% YoY

Source: Cisco GCI 2016, UN, NetCraft, Synergy Research.

Although end-user security concerns exist, the time

of amateur hackers is long over, and hacking is now
an organized crime or state-sponsored event. DDoS
attacks against customers remain a major operational
threat to service providers. Attacks against
infrastructure continue to grow in prominence.
Phishing and malware threats occur on a daily basis.

http://cybersecurityventures.com/hackerpocalypse-cybercrime-report-2016

10

https://www.skyhighnetworks.com/cloud-report/

2016 Cisco and/or its affiliates. All rights reserved.

16

Cisco Global Cloud Index: Forecast and Methodology, 20152020

White Paper
Cisco Public

According to the Cisco 2016 Annual Security report, the government has emerged as the number one high
risk industry for web malware encounters in 2015. To track high-risk verticals for web malware encounters,
the relative volumes of attack traffic (block rates) and normal or expected traffic was examined. From
January to March 2015, government was the vertical with the highest block rate activity. From March to
May, it was electronics. In midsummer, professional services saw the most blocks. And in the fall of 2015,
healthcare was leading all verticals in the number of block rates, per Figure 30.
Figure 30. Vertical Block Rates by Month, November 2014September 2015

Source: Cisco Security Research.

Encrypted traffic, particularly HTTPS, has reached

a tipping point. While not yet the majority of
transactions, it will soon become the dominant form
of traffic on the Internet. It consistently represents
over 50 percent of bytes transferred (Figure 31) due
to the HTTPS overhead and larger content that is sent
via HTTPS, such as transfers to file storage sites.
Figure 31. SSL/HTTPS Percentages
Percentage of Traffic
60

% Total Bytes

57%

46%
40
% HTTPS Requests
33.56%

20

24%
Jan.

2015

Source: Cisco Security Research.

2016 Cisco and/or its affiliates. All rights reserved.

Oct.

IoE and big data requirements are starting a new

wave of security discussions and technology
convergence. As enterprises and service providers
move to public and private clouds and modernize
data centers with SDN or consume IT as a Service
(ITaaS), security becomes an even more complex
concern. Besides hardware appliances, virtual
machines, and server software, innovative services
that use SDN and NFV will help to improve the data
integrity and security of cloud infrastructures.

Network Speeds and Latency Analysis

The cloud-readiness study offers a regional view

of the requirements for broadband and mobile
networks to deliver next-generation cloud services.
The enhancements and reliability of these networks
will support the increased adoption of cloud
computing solutions that deliver basic as well
as advanced application services. For example,

17

Cisco Global Cloud Index: Forecast and Methodology, 20152020

White Paper
Cisco Public

consumers expect to be able to communicate

with friends as well as stream music and videos
at any time, any place. Business users require
reliable access to business communications along
with mobile solutions for video conferencing
and mission-critical customer and operational
management systems.
The study also explores the ability of each global
region (Asia Pacific, Central and Eastern Europe,
Latin America, Middle East and Africa, North
America, and Western Europe) to support a
sample set of basic, intermediate, and advanced
business and consumer cloud applications. Each
regions cloud readiness is assessed with relation

to the sample services based on download and

upload fixed and mobile network speeds as well
as associated network latencies (segmented by
business and consumer connections). Download
and upload speeds as well as latencies are essential
measures to assess network capabilities for
cloud readiness. Figure 32 provides the business
and consumer cloud service categories and the
corresponding network requirements used for this
study. Tables 3 through 5 describe the requirements
and define a sample set of applications from each of
the readiness categories. Note that the concurrent
use of applications can further influence the user
experience and cloud accessibility.

Figure 32. Sample Business and Consumer Cloud Service Categories

Basic Cloud Apps

Intermediate Cloud Apps

Advanced Cloud Apps

Network Requirements:

Network Requirements:

Network Requirements:

Download Speed:
Up to 750 kbps

Download Speed:
751-2,500 kbps

Download Speed:
Higher than 2,500 kbps

Upload Speed:
Up to 250 kbps

Upload Speed:
251-1,000 kbps

Upload Speed:
Higher than 1,000 kbps

Latency: Above 160 ms

Latency: 159-100 ms

Latency: Less than 100 ms

Table 3.

Sample Basic Applications

Apps

Definitions

Download

Upload

Latency

Stream basic video

and music

Deliver sound and video without the need

to download files of different audio or video
formats using computer servers connected
to the Internet to access information.

High

Low

Medium

Text communications
(email and instant
messaging)

A cross-platform messaging application

that allows the exchange of messages
without having to pay for Short Message
Service (SMS), using an Internet data plan.

Low

Low

Medium

Voice over IP (VoIP)

(Internet telephony)

A broad range of services transmitting

voice over the Internet.

Low

Low

Medium

Web browsing

Accelerate web experiences and searching

through cloud computing using technology
to shift the workload to the cloud servers.

Low

Low

Medium

2016 Cisco and/or its affiliates. All rights reserved.

18

Cisco Global Cloud Index: Forecast and Methodology, 20152020

White Paper
Cisco Public

Apps

Definitions

Download

Upload

Latency

Web conferencing

A cloud application used to interact with

other participants and have that live
and in-person feeling for attendees; it
offers services such as collaborative web
browsing and application sharing.

Medium

Medium

Medium

Cloud-based
learning
management system

This app provides the user with the

flexibility of being able to access and
collaborate with others in a centralized
environment. With information housed in a
virtual storage environment, it allows work
to be completed outside the boundaries of
the formal learning or work institutions.

High

Medium

Medium

Table 4.

Sample Intermediate Applications

Apps

Definitions

Download

Upload

Latency

Enterprise resource
planning (ERP) and
customer relationship
management (CRM)

ERP and CRM systems allow businesses

to manage their business and business
relationships and the data and information
associated with them.

Medium

Low

Medium

High-definition (HD)
video streaming

Deliver HD video without the need to

download files of HD video formats using
computer servers connected to the Internet
to access information.

High

Low

Low

Augmented reality
(AR) gaming
applications

Augmented reality (AR) games involve a

live direct or indirect view of a physical,
real-world environment whose elements
are augmented (or supplemented) by
computer-generated sensory input such as
sound, video, graphics, or GPS data.

High

Medium

Low

Web electronic
health records
(EHRs)

EHRs are designed to contain and share

information from all providers involved
in a patients care in a structured format
allowing patient information to be easily
retrieved and transferred in ways that can
aid patient care.

Medium

High

Low

Voice over LTE

(VoLTE)

This standardized system allows for

transferring traffic for VoLTE.

Low

Low

Low

Personal content
locker

Asynchronous storage enables applications

that use compound files to efficiently
render their content when accessed by
means of existing Internet protocols, with
a single request to a server triggering the
download of nested objects contained
within a webpage, eliminating the need to
separately request each object.

High

High

Low

2016 Cisco and/or its affiliates. All rights reserved.

19

Cisco Global Cloud Index: Forecast and Methodology, 20152020

White Paper
Cisco Public

Table 5.

Sample Advanced Applications

Apps

Definitions

Download

Upload

Latency

Telemedicine

Telemedicine is the use of medical

information exchanged from one site to
another through electronic communications
to improve a patients clinical health status
and includes using two-way video, email,
smartphones, wireless tools, and other
forms of telecommunications technology.

Medium

Medium

Low

HD video
conferencing

Two-way interactive HD video

communication is delivered using Internet
technologies that allow people at different
locations to come together for a meeting.

High

High

Low

Ultra HD video
streaming

This app delivers Ultra HD video without the

need to download files of different video
formats using computer servers connected
to the Internet to access information.

High

High

Low

Virtual reality (VR)

streaming

Streaming of realistic and immersive

simulation of a three-dimensional
environment, created using interactive
software and hardware, and experienced
or controlled by movement of the body or
as an immersive, interactive experience
generated by a computer.

High

High

Low

High-frequency
stock trading

These apps support the rapid turnover of

positions through the use of sophisticated
trading algorithms, which process hundreds
of trades in fractions of a second on the
basis of changing market conditions.

Low

Low

Low

Connected vehicles
safety applications

These apps involve the development

and deployment of a fully connected
transportation system that makes the
most of multimodal, transformational
applications requiring a combination of
well-defined technologies, interfaces, and
processes that, combined, help ensure
safe, stable, interoperable, reliable system
operations that minimize risk and maximize
opportunities.

Low

Low

Low

Regional network performance statistics were ranked by their ability to support these three cloud service
categories. More than 300 million records from Ooklas Speedtest11 along with data from Ovum-Informa,
Synergy Research, Point Topic, United Nations (UN), World Bank, NetCraft, International Telecommunication
Union (ITU), International Labor Organization (ILO), and other sources were analyzed from more than 200
countries to understand cloud readiness. The regional averages of these measures are included as follows
and in Appendix G.
11

Measured by Speedtest.net, small binary files are downloaded and uploaded between the web server and the client to
estimate the connection speed in kilobits per second (kbps).

2016 Cisco and/or its affiliates. All rights reserved.

20

Cisco Global Cloud Index: Forecast and Methodology, 20152020

White Paper
Cisco Public

The cloud readiness characteristics follow.

Network Access
Internet ubiquity: This indicator measures fixed
and mobile Internet penetration while considering
population demographics to understand the
pervasiveness and expected connectivity in
various regions.
Network Performance
Download speed: With increased adoption of
mobile and fixed bandwidth-intensive applications,
end-user download speed is an important
characteristic. This indicator will continue to be
critical for the quality of service delivered to virtual
machines, CRM, and ERP cloud platforms for
businesses, video download, and content-retrieval
cloud services for consumers.
Upload speed: With the increased adoption of
virtual machines, tablets, and video conferencing in
enterprises as well as by consumers on both fixed
and mobile networks, upload speeds are especially
critical for delivery of content to the cloud. The
importance of upload speeds will continue to
increase over time, promoted by the dominance of
cloud computing and data center virtualization, the
need to transmit many millions of software updates
and patches, the distribution of large files in virtual
file systems, and the demand for consumer cloud
game services and backup storage.
Network latency: Delays experienced with VoIP,
viewing and uploading videos, online banking on
mobile broadband, or viewing hospital records
in a healthcare setting are due to high latencies
(usually reported in milliseconds). Reducing delay in
delivering packets to and from the cloud is crucial to
delivering todays advanced services (and making
sure of a high-quality end-user experience).

Global Average Download and Upload Speed

Overview (2016)

Download and upload speeds as well as latencies

are important measures to assess network
capabilities for cloud readiness. The Cisco GCI
Supplement provides additional country-level details
for download speeds, upload speeds, and latencies.
To support cloud services and applications, the
quality of the broadband connection is critical.
Although theoretical speeds offered by fixed
and mobile operators can seem adequate, many
extraneous factors are involved in the actual
network measurements. Speeds and latencies vary

2016 Cisco and/or its affiliates. All rights reserved.

within each country and region, based on urban and

rural deployment of fixed and mobile broadband
technology, proximity to traditional and cloud data
centers, and the quality of customer premises
equipment (CPE).
Lesser variability in download speeds, upload
speeds, and latency will allow consumers to access
advanced cloud applications consistently throughout
the country. To measure this variability, we have also
included the median download speeds and median
upload speeds, along with the update to the mean
or average download speeds and upload speeds, all
measured and typically represented in kilobits per
second (kbps) or megabits per second (Mbps).

Key Results

The global average fixed download speed is

29.5 Mbps, and the global median fixed download
speed is 19.1 Mbps.
The global average fixed upload speed is
14.3 Mbps, and the global median upload speed
is 6.8 Mbps.
The global average mobile download speed
is 15.1 Mbps, and the global median mobile
download speed is 8.3 Mbps.
The global average mobile upload speed is
7.6 Mbps, and the global median mobile upload
speed is 3.5 Mbps.

Regional Fixed Download and Upload Speeds

Average fixed download speeds: Asia Pacific
leads with 33.9 Mbps, and North America follows
with 32.9 Mbps.
Average fixed upload speeds: Central and
Eastern Europe leads with 19.3 Mbps, and Asia
Pacific follows with 18.9 Mbps (Figure 33). For
further details, refer to Appendix G and the Cisco
GCI Supplement.
Median fixed download and upload speeds:
Median speeds are lower than the average/mean
speeds, as shown in Figure 33, because of a
higher distribution of speeds in the region that
are lower than the mean. Besides the required
network characteristics for advanced cloud
application, for an optimal end-user experience
in larger user bases with cloud services, the
majority of speeds must also be closer to
the mean. This factor is a critical factor. To
understand speed distribution patterns in detail
for a select list of countries, refer to the Cisco
GCI Supplement.

21

Cisco Global Cloud Index: Forecast and Methodology, 20152020

White Paper
Cisco Public

Figure 33. Regional Average Fixed Speeds, 2016

2nd
Mean
DL
32.9
Mbps

1st
Median
DL
23.4
Mbps

3rd
Mean
UL
11.6
Mbps

3rd
Median
UL
5.9
Mbps

North America

5th
Mean
DL
9.3
Mbps

5th
Median
DL
6.2
Mbps

3rd
Mean
DL
30.2
Mbps

3rd
Median
DL
18.0
Mbps

4th
Mean
UL
11.0
Mbps

4th
Mean
UL
4.8
Mbps

Western Europe

6th
Mean
UL
3.3
Mbps

6th
Median
UL
1.0
Mbps

Latin America

6th
Mean
DL
7.8
Mbps

6th
Median
DL
5.7
Mbps

4th
Mean
DL
29.1
Mbps

4th
Median
DL
16.2
Mbps

1st
Mean
UL
19.3
Mbps

1st
Median
UL
10.7
Mbps

2nd
Mean
UL
19.0
Mbps

2nd
Median
UL
8.8
Mbps

Central and Eastern Europe

5th
Mean
UL
3.9
Mbps

5th
Median
UL
1.4
Mbps

Middle East and Africa

1st
Mean
DL
33.9
Mbps

2nd
Median
DL
22.4
Mbps

Asia Pacific

Source: Cisco Global Cloud Index, 20152020.

Regional Average Mobile Download and Upload Speeds

Average mobile download speeds: Asia Pacific leads with 18.5 Mbps, and Western Europe follows with
18.2 Mbps.
Average mobile upload speeds: North America leads with 9.9 Mbps, and Asia Pacific follows with 8.9
Mbps (Figure 34). For further details, refer to Appendix G and the Cisco GCI Supplement.
Median mobile download and upload speeds: Median speeds are lower than mean mobile speeds within
all regions, with the distribution of speeds in the regional population tending to be lower than the average.
Figure 34. Regional Average Mobile Speeds, 2016

3rd
Mean
DL
17.7
Mbps

3rd
Median
DL
9.6
Mbps

1st
Mean
UL
9.9
Mbps

2nd
Median
UL
4.8
Mbps

North America

5th
Mean
DL
8.4
Mbps

5th
Median
DL
4.8
Mbps

2nd
Mean
DL
18.2
Mbps

2nd
Median
DL
12.8
Mbps

3rd
Mean
UL
7.9
Mbps

3rd
Median
UL
4.0
Mbps

Western Europe

5th
Mean
UL
4.1
Mbps

6th
Median
UL
1.5
Mbps

Latin America

6th
Mean
DL
6.6
Mbps

6th
Median
DL
4.4
Mbps

Middle East and Africa

4th
Mean
DL
11.0
Mbps

4th
Median
DL
7.5
Mbps

4th
Mean
UL
6.8
Mbps

4th
Median
UL
2.5
Mbps

2nd
Mean
UL
8.9
Mbps

1st
Median
UL
5.2
Mbps

Central and Eastern Europe

6th
Mean
UL
3.9
Mbps

5th
Median
UL
2.1
Mbps

1st
Mean
DL
18.5
Mbps

1st
Median
DL
14.2
Mbps

Asia Pacific

Source: Cisco Global Cloud Index, 20152020.

Network Latency

Global average fixed latency is 36 ms.

Asia Pacific leads in average fixed latency with 26 ms, followed by Central and Eastern Europe with 30 ms.
Global average mobile latency is 81 ms.
Western Europe leads in average mobile latency with 57 ms, followed by Asia Pacific with 73 ms.

2016 Cisco and/or its affiliates. All rights reserved.

22

Cisco Global Cloud Index: Forecast and Methodology, 20152020

White Paper
Cisco Public

As cloud data centers are built and distributed around the world and content is closer to the user, the length of
the time it takes for a small packet of data to be sent and received will get lesser. The other factor that can also
affect latency is congestion, which leads to a lower throughput. Latency has significantly improved in both fixed
and mobile networks. Figures 35 and 36 show the latency improvements from 2014 through 2016 in average
fixed latency in ms as well as average mobile latency in ms by region.
Figure 35. Improvements in Average Fixed Latency in ms, 20142016
87
77
62
Middle East
and Africa

Asia Pacific
40

35
26

Central and
Eastern
Europe

49

North
America

47
33

69

42

38

30

64
54

Latin
America

2014

2015

Western
Europe

46

2016

44

38

2015

2014

2016

Source: Cisco Global Cloud Index, 20152020.

Figure 36. Improvements in Average Mobile Latency in ms, 20142016

Asia Pacific 2014
2015
2016
Central and 2014
Eastern
2015
Europe
2016
Latin
America

2014
2015
2016

Middle East 2014

and Africa
2015
2016
North
America

2014
2015
2016

Western
Europe

2014
2015
2016
0

20

40

60

80

Source: Cisco Global Cloud Index, 20152020.

2016 Cisco and/or its affiliates. All rights reserved.

100

120

140

160
180
Latency (ms)

200

220

240

260

280

300

320

340

23

Cisco Global Cloud Index: Forecast and Methodology, 20152020

White Paper
Cisco Public

For further details, refer to Appendix G and the

Cisco GCI Supplement.

Conclusion
In summary, we can draw several main conclusions
from the updated Cisco GCI 20152020 report.
Global data center traffic is firmly in the zettabyte
era and will more than triple from 2015 to reach
15.3 ZB annually by 2020. Not only is the data
center traffic growing, but it is also getting
streamlined with architectural innovations such as
SDN and NFV, which offer new levels of optimization
for data centers. A rapidly growing segment of
data center traffic is cloud traffic, which will nearly
quadruple over the forecast period and represent
92 percent of all data center traffic by 2020. An
important traffic enabler in the rapid expansion
of cloud computing is increasing data center
virtualization, which provides services that are
flexible, fast-to-deploy, and efficient. By 2020, 92
percent of all workloads will be processed in the
cloud.
Within the cloud segment private cloud will have
significantly more workloads than the public cloud
initially, but public cloud will grow faster than the
private cloud over the forecast period, and by 2016
the majority share of workloads will transition to
public cloud. This will also be depicted in the degree
of virtualization: the workload density in public cloud
will outpace that in private cloud by 2016 as well.
As the business sensitivity to costs associated
with dedicated IT resources grows along with
demand for agility, we can see a greater adoption
of public cloud by the businesses, especially
with strengthening of public cloud security. Many
enterprises will adopt a hybrid approach to cloud
as they transition some workloads from internally
managed private clouds to externally managed
public clouds. All three types of cloud service
delivery modelsIaaS, PaaS, and SaaSwill continue
to grow as more and more businesses realize the
benefits of moving to a cloud environment.
Additional trends influencing the growth of data
center and cloud computing include increasing
digitization, the widespread adoption of multiple
devices and connections or the IoE phenomenon,
and the growth of mobility. An extraordinary amount

2016 Cisco and/or its affiliates. All rights reserved.

of data is being generated by IoE applicationsto

the tune of 600 ZB by 2020. However, only a
relatively very small portion of that content (about
6.2 ZB), will be stored. Over time, more and more
of the data resident on client devices will move to
the data center. Total data center storage capacity
will grow nearly 5-fold from 2015 to 2020, growing
from 352 EB in 2015 to 1.8 ZB by 2020. Cloud
will accounts for nearly 90 percent of the installed
storage capacity by 2020. To address rising user
demands, cloud-based services such as consumer
cloud storage are gaining momentum. By 2020,
nearly 60 percent (59%) of the consumer Internet
population will be using personal cloud storage.
Growth in M2M connections and applications is also
driving new data analytics needs.
This study also considers the importance of cloud
readiness. Based on the analysis of regional average
download and upload speeds and latencies for
business and consumer mobile and fixed networks,
all regions have made significant strides to reach a
capable level of supporting basic and intermediate
cloud services. The focus now turns to continuing
to improve network capabilities to support the
advanced cloud applications that organizations and
end users expect and rely upon.

For More Information

For more information, visit www.cisco.com/go/
cloudindex.

Appendix A: Data Center Traffic Forecast

Methodology
Figure 37 outlines the methodology used to forecast
data center and cloud traffic. The methodology
begins with the installed base of workloads
categorized by workload type and implementation
and then applies the volume of bytes per workload
per month to obtain the traffic for current and future
years.
Figure 37. Data Center Traffic Forecast Methodology
Installed Base
of Workloads

Bytes of Traffic
per Workload
per Month

Percent Traffic
Within Data Center
and Data Center
to Data Center

Analyst Data

Measured Data

Measured Data

= Traffic

24

Cisco Global Cloud Index: Forecast and Methodology, 20152020

White Paper
Cisco Public

Analyst Data

Data from several analyst firms and international

agencies (including Gartner, IDC, Juniper Research,
Ovum, Synergy, ITU, and the United Nations) was
used as inputs to the Global Cloud Index analysis. For
example, analyst data was considered to calculate
an installed base of workloads by workload type and
implementation (cloud or noncloud). The analyst input
consisted of server shipments with specified workload
types and implementations. Cisco then estimated the
installed base of servers and the number of workloads
per server to obtain an installed base of workloads.

Measured Data

Network data was collected from a variety of

enterprise and Internet centers. The architectures
of the data centers analyzed vary, with some
having a three-tiered and others a two-tiered
architecture. For three-tiered data centers, data
was collected from four points: the link from the
access routers to the aggregation routers, the link
from the aggregation switches or routers to the site
or regional backbone router, the WAN gateway, and
the Internet gateway. For two-tiered data centers,
data was collected from three points: the link from
the access routers to the aggregation routers, the
WAN gateway, and the Internet gateway.
For enterprise data centers, any traffic measured
northbound of the aggregation also carries nondata
center traffic to and from the local business campus.
For this reason, to obtain ratios of the volume of
traffic carried at each tier, it was necessary to
measure the traffic by conversations between hosts
rather than traffic between interfaces, so that the
nondata center conversations could be eliminated.
The hosts at either end of the conversation were
identified and categorized by location and type. To
be considered data center traffic, at least one of the
conversation pairs had to be identified as appearing
in the link between the data center aggregation
switch or router and the access switch or router.
In addition, as noted in this paper, the methodology
for the estimation of cloud data center traffic has
changed since the last release of the Cisco Global
Cloud Index. The previous methodology included
all storage traffic in the noncloud traffic category.
The updated methodology includes storage traffic
associated with cloud workloads in the cloud traffic
category. For example, storage traffic associated with
cloud application development would be counted as
cloud traffic in the updated methodology, but would
have been excluded in the previous methodology.

2016 Cisco and/or its affiliates. All rights reserved.

Appendix B: Global Cloud Index and Visual

Networking Index
The Cisco Global Cloud Index (GCI) and Cisco Visual
Networking Index (VNI) are distinct forecasts that
have an area of overlap. The Cisco VNI forecasts
the amount of traffic crossing the Internet and IP
WAN networks, whereas the Cisco GCI forecasts
traffic within the data center, from data center to
data center, and from data center to user. The Cisco
VNI forecast consists of data center-to-user traffic,
along with nondata center traffic not included in the
Cisco GCI (various types of peer-to-peer traffic).
The Cisco GCI includes data-centerto-user traffic
(this area is the overlap with the Cisco VNI) datacentertodata center traffic, and traffic within the
data center. The Cisco VNI forecasts the amount of
traffic crossing the Internet and IP WAN networks
(Figure 38).
Figure 38. Cisco VNI and Global Cloud Index
Visual Networking Index (VNI)
A

2.3 ZBs

Global Cloud Index (GCI)

D

11.7 ZBs

Non-Data Center

1.4 ZBs

NOT included in GCI

2.2 ZBs

Data Center-to-User

0.1 ZBs

A Traffic

B Traffic

This is the overlap

between VNI
and GCI

B
B

15.3 ZBs

Data Center-to-User
Traffic (14%)
This is the overlap
between VNI and GCI
Data Center-to-Data

C Center Traffic (9%)

Traffic that flows from

data center to data center
Within Data

D Center (77%)
Source: Cisco Global Cloud Index, 20152020.

Traffic that remains

within the data center

Appendix C: Regional Cloud Traffic Trends

The Cisco Global Cloud Index includes regional
forecast data for cloud traffic growth (Table 6).
In 2015, North America generated the most cloud
traffic (1,891 EB annually), followed by Asia Pacific
(908 EB annually) and Western Europe (718 EB
annually).
By 2020, North America will generate the most
cloud traffic (3.6 ZB annually), closely followed by
Asia Pacific (2.3 ZB annually) and Western Europe
(1.5 ZB annually).
From 2015 to 2020, the Middle East and Africa is
expected to have the highest cloud traffic growth
rate (41 percent CAGR), followed by Central and
Eastern Europe (38 percent CAGR) and North
America (33 percent CAGR).

25

Cisco Global Cloud Index: Forecast and Methodology, 20152020

White Paper
Cisco Public

Table 6.

Cloud Traffic Growth by Region, in EB

Region

2015

2016

2017

2018

2019

2020

CAGR
201520

Asia Pacific

908

1,367

1,871

2,387

2,923

3,469

31%

Central and Eastern Europe

124

168

221

291

379

485

31%

Latin America

140

190

242

304

371

448

26%

Middle East and Africa

69

105

145

191

242

304

34%

North America

1,891

2,771

3,838

4,860

5,779

6,844

29%

Western Europe

718

1,035

1,395

1,768

2,157

2,528

29%

Source: Cisco Global Cloud Index, 20152020.

Appendix D: Workload Distribution by Region

Tables 7, 8, and 9 summarize data center workloads by type and region.
Table 7.

Regional Distribution of Total Data Center Workloads, in Millions

Total Data Center Workloads in Millions

2015

2016

2017

2018

2019

2020

CAGR
20152020

Asia Pacific

48.6

68.1

91.3

115.7

39.7

163.4

27.4%

Central and Eastern Europe

5.2

6.6

8.3

10.0

11.6

12.9

20.1%

Latin America

6.5

8.4

10.6

13.0

15.1

17.1

21.3%

Middle East and Africa

4.7

6.1

7.8

9.5

11.0

12.4

21.6%

North America

78.5

99.1

124.1

149.3

170.8

189.1

19.2%

Western Europe

37.4

46.6

57.5

67.9

76.6

83.9

17.5%

Source: Cisco Global Cloud Index, 20152020.

Table 8.

Regional Distribution of Cloud Workloads, in Millions

Cloud Data Center Workloads in Million

2015

2016

2017

2018

2019

2020

CAGR
20152020

Asia Pacific

38.3

57.0

79.6

103.6

127.6

151.5

31.6%

Central and Eastern Europe

4.0

5.5

7.3

9.1

10.7

12.2

25.1%

Latin America

5.2

7.2

9.6

12.0

14.1

16.2

25.5%

Middle East and Africa

3.6

5.1

6.9

8.7

10.3

11.8

26.6%

North America

57.2

77.9

103.6

129.5

152.3

172.1

24.6%

Western Europe

27.7

37.1

48.4

59.1

68.4

76.3

22.5%

Source: Cisco Global Cloud Index, 20152020.

2016 Cisco and/or its affiliates. All rights reserved.

26

Cisco Global Cloud Index: Forecast and Methodology, 20152020

White Paper
Cisco Public

Table 9.

Regional Distribution of Traditional Data Center Workloads, in Millions

Traditional Data Center Workloads in Millions

2015

2016

2017

2018

2019

2020

CAGR
20152020

Asia Pacific

10.3

11.1

11.7

12.1

12.1

11.9

2.9%

Central and Eastern Europe

1.2

1.1

1.0

0.9

0.9

0.8

-8.5%

Latin America

1.3

1.2

1.1

1.0

1.0

0.9

-7.4%

Middle East and Africa

1.0

1.0

0.9

0.8

0.7

0.6

-10.3%

North America

21.3

21.2

20.5

19.8

18.6

17.0

-4.4%

Western Europe

9.7

9.5

9.1

8.7

8.3

7.6

-4.8%

Source: Cisco Global Cloud Index, 20152020.

Appendix E: Cloud Definitions

Cloud Definition

The Cisco GCI aligns with the industry-standard

cloud computing definition from the National
Institute of Technology (NIST). The NIST definition
lists five essential characteristics of cloud computing
(Figure 39):

of software, platform, and infrastructure services.

Cloud data centers can be operated by service
providers as well as private enterprises.

Broad network access

However, there is a slight variation from the

NIST definition on how we classify private and
public clouds. A cloud service could be public or
private, depending on the demarcation linethe
physical or virtual demarcationbetween the public
telecommunications network (WAN) and the private
network of an organization (LAN) (Figure 40).

Resource pooling

Figure 40. Cloud Deployment Models

On-demand self-service

Rapid elasticity or expansion

Measured service

Enterprise Network

Service Provider Network

Figure 39. Essential Characteristics of Cloud

Measured
Service

On Demand/
Self Service

Cloud
Rapid
Elasticity

Broad
Network
Access

Cloud deployment models include private, public,

and hybrid clouds (or a combination of them). These
distinct forms of cloud computing enable a variety

2016 Cisco and/or its affiliates. All rights reserved.

Hybrid Cloud

E.g. Cisco Cloud owned

and managed by Cisco for
its own employees,
customers and partners.

Resource
Pooling

Cloud Deployment Models

Private
Cloud

Public
Cloud

E.g. AT&T, Verizon,

Amazon AWS, Microsoft
Azure, Salesforce, Google.

If the cloud assets lie on the service provider side

of the demarcation line, then it would be considered
a public cloud service. Virtual private cloud (VPC)
falls in this category. Also the multitenant consumer
cloud services would fall in this category.
If the cloud assets lie on the organization side of
the demarcation line, then the service would be
considered a private cloud service. In general,
a dedicated cloud owned and managed by an
organizations IT would be considered a private cloud.

27

Cisco Global Cloud Index: Forecast and Methodology, 20152020

White Paper
Cisco Public

Hybrid cloud, as the name suggests, is a

combination of public and private clouds. In a hybrid
cloud environment, some of the cloud computing
resources are managed in-house by an enterprise
and some are managed by an external provider.
We define private and public as distinct categories;
we do not separately break out the hybrid cloud
because it is simply a superset of the private and
public clouds in varying degrees.

Cloud Service Models (IaaS, PaaS, and SaaS)

The Cisco GCI forecast for cloud workload splits

across the three main cloud services models: SaaS,
PaaS, and IaaS (Figure 41). They are defined in line
with NISTs definitions.
Figure 41. Cloud Service Models: IaaS, PaaS, and SaaS
Software as a Service
(SaaS)

Platform as a Service
(PaaS)

Infrastructure as a
Service (IaaS)

Applcations

Applcations

Applcations

Data

Data

Data

Middleware

Middleware

Middleware

Operating System

Operating System

Operating System

Virtualization

Virtualization

Virtualization

Servers

Servers

Servers

Storage

Storage

Storage

Networking

Networking

Networking

Cloud Customer
Manages

Cloud Provider Manages

SaaS

The capability provided to the consumer is to use

the providers applications running on a cloud
infrastructure. The applications are accessible from
various client devices through either a thin-client
interface, such as a web browser (for example,
web-based email) or a program interface. The
consumer neither manages nor controls the
underlying cloud infrastructure, including networks,
servers, operating systems, storage, or even
individual application capabilities, with the possible
exception of limited user-specific application
configuration settings.

PaaS

The capability provided to the consumer is to deploy

onto the cloud infrastructure consumer-created or
-acquired applications created using programming
languages, libraries, services, and tools supported

2016 Cisco and/or its affiliates. All rights reserved.

by the provider. The consumer neither manages

nor controls the underlying cloud infrastructure,
including network, servers, operating systems,
or storage, but has control over the deployed
applications and possibly configuration settings for
the application-hosting environment.

IaaS

The capability provided to the consumer is to

provision processing, storage, networks, and
other fundamental computing resources where
the consumer is able to deploy and run arbitrary
software, which can include operating systems
and applications. The consumer neither manages
nor controls the underlying cloud infrastructure but
has control over operating systems, storage, and
deployed applications; and possibly limited control
of select networking components (for example, host
firewalls).

Appendix F: Workload Application Definitions

Following is the list of application definitions used for
segmenting workloads in trend 4:
Enterprise Workload Applications
Compute: mainly covers cloud IaaS
Collaboration: email, conferencing, enterprise
social networking, file sharing, content
management
Database/analytics/IoE: on-premises database
apps, big data apps, business intelligence, dataoriented PaaS services
Other ERP and enterprise apps: CRM, HCM,
finance apps, storage, service/systems/
operations management, and so on
Consumer Workload Applications
Search: search
Social Networking: Facebook, Twitter, Tencent,
LinkedIn, Google+, Snapchat, and so on
Video/streaming media: video sharing and video
streaming services
Other consumer apps: email, messaging, storage,
file sharing, music services, e-commerce, news,
and so on

28

Cisco Global Cloud Index: Forecast and Methodology, 20152020

White Paper
Cisco Public

Appendix G: Regional Cloud Readiness Summary

Table 10 summarizes cloud readiness for businesses and consumers by region, considering download and
upload speeds, and latency. For more details, refer to the Cisco GCI Supplement.
Table 10. Regional Cloud Readiness

Network

Region

Average Download
Speeds (Mbps)

Average Upload
Speeds (Mbps)

Average Latency
(ms)

Fixed

Asia Pacific

28.1

15.9

35

Central and Eastern Europe

28.3

20.9

33

Latin America

7.6

2.4

64

Middle East and Africa

7.0

2.2

77

North America

25.4

8.8

42

Western Europe

22.8

7.4

44

Asia Pacific

12.1

6.1

83

Central and Eastern Europe

10.9

7.7

75

Latin America

5.7

2.0

116

Middle East and Africa

4.5

2.3

156

North America

16.3

6.5

63

Western Europe

13.7

4.8

70

Mobile

Source: Cisco Global Cloud Index, 20152020.

2016 Cisco and/or its affiliates. All rights reserved. Cisco and the Cisco logo are trademarks or registered trademarks of Cisco and/or its affiliates in the U.S.
and other countries. To view a list of Cisco trademarks, go to this URL: www.cisco.com/go/trademarks. Third-party trademarks mentioned are the property of
their respective owners. The use of the word partner does not imply a partnership relationship between Cisco and any other company. (1110R)
C11-738085-0011/16