The term “cloud” refers to the
way IT resources are configured and
accessed, not the size of the facility.
A very wide range of data centers are
moving to cloud computing. The
accelerating growth of the cloud
can also be viewed through the
prism of private clouds compared
to public clouds and non-cloud data
centers, as shown in Figure 7.
The focus tends to be on public
cloud data centers since these data
centers have the highest growth
rate and tend to be operated by
the largest technology companies.
However, private cloud data centers
are also projected to achieve doubledigit
compound annual growth rates.
Private cloud data centers can
be on-premises or off-premises
52 I ICT TODAY
Global Private vs. Public vs. Non-Cloud
Enterprise Application Workloads and Compute Instances
by Google,12 most employ a mix of copper, short reach
(SR) multimode optics, and long reach (LR) or coarse
wavelength division multiplexing (CWDM4/CLR4)
singlemode optics.13 The most critical metrics for
intra-datacenter interconnects are cost, size, and
power in order of importance.14
Other valid definitions for “hyperscale” primarily
refer to the size of the physical facility and the long cable
runs required (e.g., reaches of more than 300 m and up
to 2 km). The term “mega data center” is also often used
to define such installations.14
Traditional Enterprise Data Centers
The rapid growth of cloud data centers and the decline
of traditional data centers often leads to the misinterpretation
that enterprise data centers are disappearing.
In fact, a significant percentage of large enterprise data
centers are being transformed to cloud data centers. They
are utilizing server virtualization, containerization,
software defined networks (SDN), network functions
virtualization (NFV), analytics and automation to provide
cloud computing capabilities inside the company for
rapid on-demand self-service, broad network access,
and measured service.
450
400
350
300
250
200
150
100
50
(e.g., multitenant data centers, colocation data centers).
Non-cloud data centers are the only part of the market
that is shrinking.
Hyperscale Data Centers
The Cisco GCI defines a hyperscale data center based
on meeting one of the revenue criteria:9
• More than US $1 billion in annual revenue
from IaaS, PaaS or infrastructure hosting services
• More than US $2 billion in annual revenue
from SaaS
• More than US $4 billion in annual revenue
from Internet, search and social networking
• More than US $8 billion in annual revenue
from e-commerce/payment processing
Twenty-four providers meet the Cisco GCI hyperscale
criteria. These 24 hyperscale data center companies utilize
a wide variety of architectures and both MMF and SMF
optical links. Some of the largest U.S. and China-based
hyperscale cloud companies are currently deploying
MMF-based 100G-SR4 or 400G-SR8.10 Others have
adopted parallel SMF 4-lane (PSM4) or a duplex SMF
CWDM solution.11 According to a 2017 OFC presentation
2016
0
2017 2018 2019 2020 2021
Installed Workloads & Compute
Instances in Millions
Public Cloud Data Center (32% CGAR)
Private Cloud Data Center (11% CGAR)
Non-Cloud Data Center (-5% CGAR)
57%
32%
46%
22%
35%
8%
FIGURE 7: Private versus public versus non-cloud data centers. Source: Cisco Global
Cloud Index, 2016-2021.