Chapter 1

Overview of Cloud Computing Technology

By David Ma1

I. Introduction
Since its introduction, cloud computing has rapidly grown to become an important com-
ponent of the information technologies that businesses use, and although spending on
cloud computing constitutes less than five percent of overall spending on information
technology, spending on cloud computing has grown at 4.5 times the rate of general in-
formation technology spending since 2009, a figure that is projected to rise by a multiple
of six in the near future.2 A 2016 survey of chief information officers revealed that 16 per-
cent of data processing workloads run in the public cloud, which will grow to 41 percent
by 2021.3 Of the Fortune Global 50 companies, 48 have announced plans to adopt cloud
computing.4 In other words, the importance of cloud computing to businesses is likely to
continue growing.
The objective of this chapter is to provide a brief overview of the technology that
enables cloud computing, the general characteristics and types of cloud computing, the
models that can be used to deploy them, and some advantages and disadvantages of cloud
computing.5

II. Technical Elements of Cloud Computing

An oft-cited definition of cloud computing is one developed by the National Institute
of Standards and Technology (NIST): “a model for enabling ubiquitous, convenient, on-
demand network access to a shared pool of configurable computing resources (e.g., net-
works, servers, storage, applications, and services) that can be rapidly provisioned and

1 The views expressed are solely those of the author. References to vendors are for illustrative purposes
only and do not constitute a recommendation or endorsement of such vendors. The author would like to
acknowledge and thank Jerry Dawkins, CEO and founder of True Digital Security, for his helpful review
and comments on drafts of this chapter. Errors and omissions remain the sole responsibility of the author.
2 John F. Gantz & Pam Miller, The Salesforce Economy: Enabling 1.9 Million New Jobs and
$389 Billion in New Revenue Over the Next Five Years 3 (2016), https://www.salesforce.com/
content/dam/web/en_us/www/academic-alliance/datasheets/IDC-salesforce-economy-study-2016.pdf.
3 Tiernan Ray, Amazon Seeing “Momentous” Change of Guard as Public Cloud “Booms,” Says JP Mor-
gan, Barron’s, Apr. 14, 2016, http://‌www.‌barrons.‌com/‌articles/‌amazon-‌seeing-‌momentous-‌change-­‌of-
‌guard-‌as-‌public-‌cloud-‌booms-‌says-‌jp-‌morgan-‌1460669610.
4 Mark Brinda & Michael Heric, The Changing Faces of the Cloud, Bain Brief, Jan. 25, 2017, http://‌
www.‌bain.‌com/‌publications/‌articles/‌the-‌changing-‌faces-‌of-‌the-‌cloud.‌aspx.
5 This chapter summarizes only some of the more significant technologies used to facilitate cloud com-
puting. There are others not addressed here, such as those related to usage monitoring, metering and
billing, automated scaling mechanisms, load balancers, service-level monitors, and identity and security
management. For a more comprehensive list of cloud-related technologies, see the list of projects that
comprise OpenStack, https://‌www.‌openstack.‌org/‌software/.

ABA_15_Cloud_Computing_CH01.indd 1 09/04/19 3:47 PM

 2 Cloud 3.0: Drafting and Negotiating Cloud Computing Agreements

released with minimal management effort or service provider interaction.”6 Although the
NIST Definition is helpful in describing the characteristics of cloud computing, it is not
quite as helpful in explaining exactly what the cloud is.
Perhaps another useful starting point is the history of the term “cloud.” Well before
the advent of cloud computing, computer network designers used a pictogram of a cloud
in networking diagrams to identify a group of networked computers connected to the net-
work depicted.7 The purpose of using the pictogram was to keep the diagram simple—the
cloud was intended to depict the abstract notion of computing resources that are linked
to a network where the details were either unknown or not particularly important. The
notion of abstraction is particularly appropriate when it comes to cloud computing. At
the conceptual level, cloud computing can perhaps best be thought of as the abstraction of
computing resources. The following sections provide further detail on how that abstrac-
tion is achieved.

A. Hardware
Cloud computing has always been (and likely always will be) powered by computers and
related equipment in some shape or form. For cloud computing, this will typically consist
of hundreds or thousands of computers densely packed into data centers, networked to-
gether along with storage devices with connectivity to the internet.8

B. Network Connectivity
It should come as no surprise that one of the enabling technologies for cloud computing
is network connectivity. Although cloud computing services can in theory be provided
through various types of network connections, most if not all current cloud computing
services are delivered through the internet. As cloud computing often requires the rapid
transmission of significant amounts of data between the cloud computing service pro-
vider and its users, the rise of broadband and mobile internet connectivity over the past
decade has been instrumental in enabling the corresponding rise in cloud computing.9

6 Peter Mell & Timothy Grance, National Institute of Standards and Technology, The
NIST Definition of Cloud Computing (Special Publication 800-145) (2011), at 2, http://nvlpubs.nist.
gov/nistpubs/Legacy/SP/nistspecialpublication800-145.pdf [hereinafter NIST Definition].
7 See Rebecca J. Rosen, Clouds: The Most Useful Metaphor of All Time?, Atlantic, Sept. 30, 2011,
https://‌w ww.‌theatlantic.‌com/‌technology/‌archive/‌2011/‌09/‌clouds-‌the-‌most-‌useful-‌­metaphor-‌of-‌all-
‌time/‌245851/. For a more in-depth exploration of the origins of the terms, see Antonio Regalado, Who
Coined “Cloud Computing”?, MIT Technology Review, Oct. 31, 2011, https://‌www.‌technologyreview.‌com/‌
s/‌425970/‌who-‌coined-‌cloud-‌computing/.
8 Readers should be cautioned, however, that “typically” does not equate to “necessarily”—the term
“cloud” can be used generically to describe any remote computing resources, including far less robust ar-
rangements. Thus, when prospective service providers indicate that they utilize “the cloud,” users would
be well advised to enquire exactly what that entails.
9 According to the Federal Communications Commission, median download speed in the United
States quadrupled between 2011 and 2015 from 10 Mbps (megabits per second) to 41 Mbps. FCC, Mea-
suring Broadband America: Fixed Broadband Report 15 (2016), https://www.fcc.gov/reports-
research/reports/measuring-broadband-america/measuring-fixed-broadband-report-2016.

ABA_15_Cloud_Computing_CH01.indd 2 09/04/19 3:47 PM


Overview of Cloud Computing Technology  3

C. Communication Standards
The development and rapid adoption of web, internet, and related technology standards
has also played an important role in enabling the rise of cloud computing by setting out
standardized formats and protocols for receiving, transmitting, and displaying data and
information. Service providers can conform to these uniform standards, knowing that a
broad base of users can readily access them, including through browsers that are already
installed on their computers, obviating the need to install specialized software.
A subset of such standards that is particularly important for cloud computing are
the service standards and protocols that are used to enable applications on two different
computers to communicate and interact with each other.10

D. Virtualization
There is of course more to cloud computing than just computers. One of the key enabling
technologies for cloud computing is virtualization technology. This technology enables
simulation of the hardware of an actual computer in software, commonly referred to as
a “virtual machine.” A somewhat rough analogy11 is a flight simulator game: the game
simulates and presents the experience of flying an airplane, including responding and re-
acting to inputs in a manner similar to a real airplane. Virtualization software (sometimes
referred to as a “hypervisor”) does something similar, except that instead of an airplane,
it simulates (or “virtualizes”) a computer and does so very accurately—so accurately that
almost any software that can be installed on a physical computer can also be installed
and operated within that virtual machine. Virtual machines mimic, in software, all the
elements of a physical computer, including a central processing unit, memory, storage,
and networking interfaces, subject, of course, to the limitations of the physical comput-
er (the “host”) on which the hypervisor is installed. As a result, one can create a virtual
machine running the Linux operating system, for example, even though the underlying
host is running Windows.12 One can even install multiple virtual machines on a single
host, each running a different operating system (again, subject to the limits of the physical
computer). Moreover, any software running within a virtual machine is isolated within

10 These include technologies such as Simple Object Access Protocol, representational state transfer,
Web Service Description Language, and Universal Description, Discovery, and Integration, among
­others.
11 It should perhaps be emphasized that this is a conceptual rather than a technical analogy: the design
and architecture of a flight simulator bears little to no resemblance to virtualization software. That being
said, one of the earliest implementations of virtual machines on personal computers was developed spe-
cifically for a text-based adventure game called Zork, to reduce its size so that it could be run on personal
computers (as it had originally been developed on a much larger mainframe computer), and to enable
portability across different hardware platforms. See P. David Lebling, Zork and the Future of Computer-
ized Fantasy Simulations, 5 Byte 12, 172 (Dec. 1980), https://‌archive.‌org/‌details/‌byte-‌magazine-‌1980-‌12.
12 There are a number of software packages that enable one to experience virtualization firsthand
on standard desktop computers, such as VirtualBox, https://‌www.‌virtualbox.‌org/ (free and available
for Windows, Linux and Mac) or Parallels Desktop, https://‌www.‌parallels.‌com/‌ca/‌products/‌desktop
(paid, for Mac only). Microsoft Hyper-V can also be enabled on certain versions of Windows 8
and 10. Anthony Bartolo, Step-By-Step: Enabling Hyper-V for Use on Windows 10, Canitpro.net,
Sept.   8,   2015,   https://‌blogs.‌technet.‌microsoft.‌com/‌canitpro/‌2015/‌09/‌08/‌step-‌by-‌step-‌enabling-‌hyper-
‌v-‌for-‌use-‌on-‌windows-‌10/.

ABA_15_Cloud_Computing_CH01.indd 3 09/04/19 3:47 PM

 4 Cloud 3.0: Drafting and Negotiating Cloud Computing Agreements

that virtual machine and cannot communicate with or affect either the physical computer
or other virtual machines installed on that physical computer.13
Because virtual machines are created entirely in software, they provide all of the ben-
efits of software, many of which are key to enabling cloud computing. These include:
1) Portability and replicability. Given that a virtual machine is created in software,
it can be easily saved as a file and copied or moved to and operated on another
host with virtualization software installed, without the need to reinstall and re-
configure the operating system and all applications installed on the virtual ma-
chine, making it much faster and easier to move to another host, or to provision
a new server by simply copying an existing virtual machine that has already been
configured.
2) Reliability and availability. Some hypervisors can be configured to automatically
move a virtual machine to another host under certain conditions (for example, if
it detects a possible failure of the host)14 or to maintain an identical running copy
of the virtual machine on another host and automatically switch to that copy, in
both cases enhancing availability and reliability, and enabling easier disaster re-
covery.
3) Greater hardware flexibility. Given that virtualization enables virtual machines
to have characteristics independent of the hardware on which they run, there is
typically more flexibility when selecting and purchasing physical computers so
long as they are compatible with the applicable virtualization software.
4) Recovery. Given that multiple images of virtual machines can be saved at any
time, periodic “snapshots” can be created. If a virtual machine fails for any reason
(for example, after a failed upgrade or patch to the operating system), it can sim-
ply be “rolled back” to an earlier snapshot that was working properly.
5) Security and stability. Given that virtual machines are isolated from both each
other and the host computer, they can be used to enhance security, for example
by creating a separate virtual machine to run test environments for new or exper-
imental applications so that failures or crashes of such applications will impact
only that separate virtual machine.15
6) Efficiency. Physical servers are often underutilized. Virtualization can enable
greater efficiency by replacing physical computers with virtual machines and con-
solidating those virtual machines onto fewer physical computers, reducing both
hardware and operating costs.

13 This is to some extent an overgeneralization. A hypervisor can be configured to enable interaction

between a given virtual machine and its host or other virtual machines by, for example, creating virtual
network connections between them. Certain types of virtualization technology also allow for (or require)
some limited awareness of the underlying physical host within the guest operating system. In addition,
vulnerabilities occasionally are discovered that could be used to access the host system from within
a virtual machine, typically referred to as virtual machine escape. See, e.g., Dan Goodin, Extremely
Serious Virtual Machine Bug Threatens Cloud Providers Everywhere, Ars Technica, May 13, 2015, https://‌
arstechnica.‌c om/‌i nformation-technology/‌2 015/‌0 5/‌e xtremely-serious-‌v irtual-‌m achine-‌b ug-‌
threatens-‌cloud-‌providers-‌everywhere/.
14 See also Part II.F., infra.
15 Security is thereby enhanced as compared to running such applications on the same computer with-
out the use of virtualization. Security would not be improved as compared to running such applications
on a separate physical computer, although of course doing so would entail additional cost.

ABA_15_Cloud_Computing_CH01.indd 4 09/04/19 3:47 PM


Overview of Cloud Computing Technology  5

One notable variation of virtualization is containerization.16 In contrast to “full” vir-

tualization, which runs a completely separate operating system within a virtual machine,
a container utilizes some elements of the operating system installed on the host, but still
remains isolated from the host. Containers share some traits with full virtualization, such
as ease of porting or replicating. However, containers typically require significantly less
overhead than full virtualization due to their reliance on elements of the underlying op-
erating system, resulting in lower resource usage and faster creation and execution. Thus,
a given physical computer can host a much larger number of containers than virtual ma-
chines. Lastly, some container technologies allow for portability across different hosts.17

E. Multitenant Technology
“Multitenant technology” is a term used to describe an approach to application devel-
opment that enables multiple users (or multiple groups of users) to each access the same
instance of an application in such a way that the application appears to be customized for
each such user and logically isolated from every other user.
Although multitenant architecture can achieve the same result as virtualization, the
two technologies are distinct from one other. For example, a service provider could service
multiple users by creating a distinct virtual machine for each such user and installing the
same non-multitenant application in each virtual machine. Given that virtual machines are
isolated from one another, each user would have its own uniquely configured application.
In contrast, if the same application were developed using multitenant architecture, each
user18 would access a single instance of the application. The data and configuration stored
for each user would be distinct, however, and accessible only by that user through the use of
logical controls within the application. Multitenant architecture can also be combined with
virtualization. For example, a single instance of a multitenant application could be installed
and operated within a virtual machine to serve multiple users. Alternatively, there could be
multiple instances in multiple virtual machines, with each instance serving multiple users.19
Although multitenant architecture is typically much more complex (and therefore
more costly) to develop, it can result in cost savings as it enables resources to be shared
more efficiently across users, as compared to creating an individual virtual machine and
individual instance of the application for each user. However, it may reduce the flexibility
afforded to end-users. For example, given that all users of a multitenant application are
accessing the same instance of that application, they typically must all use the same ver-
sion of the application.

16 This is sometimes called container-based virtualization or operating system level virtualization.

17 This was perhaps the primary benefit of Docker (a particular type of container technology; see What
Is Docker, Docker, https://‌www.‌docker.‌com/‌what-‌docker) as compared to other container technologies,
such as Linux containers, which are to some degree dependent on the configuration of the underly-
ing host. See Chenxi Wang, What is Docker? Linux Containers Explained, InfoWorld, June 29, 2017,
https://‌www.‌infoworld.‌com/‌article/‌3204171/‌linux/‌what-‌is-‌docker-‌linux-‌containers-‌explained.html.
18 In this context, a “user” could be an organization with multiple individual users, all of whom could
access the data and configuration for that organization.
19 See, e.g., Thomas Y. Kwok et al., A Software as a Service with Multi-tenancy Support for an
Electronic Contract Management Application (IBM Research Report RC24537 (W0804-071))
(2008), http://domino.research.ibm.com/library/cyberdig.nsf/papers/B81E880A2CD5B66A852574BF-
005092C4/$File/rc23537.pdf. Although somewhat dated, this paper provides a useful overview of the
design and architecture of a multitenant SaaS application, including figures that depict the distinction
between single-tenant and multitenant approaches.

ABA_15_Cloud_Computing_CH01.indd 5 09/04/19 3:47 PM

 6 Cloud 3.0: Drafting and Negotiating Cloud Computing Agreements

F. Resource-Pooling, Allocation, and Management Technologies

Another group of technologies that facilitate cloud computing are those that are used to
pool, allocate, and manage the underlying physical resources, such as processing power,
network communications, and data storage for use by individual users. Such physical re-
sources are typically comprised of hundreds or thousands (or more) computers located in
one or multiple data centers, each connected through networking equipment.
One example of such technology is virtual infrastructure managers (VIMs). VIMs
are used to manage virtual machines across physical servers. For example, assume that a
(quite small) cloud service provider uses two computers. Each has a hypervisor installed
on it, which in turn hosts multiple virtual machines. If one computer is nearing the limits
of its capacity and a request for a new virtual machine comes in, a VIM can be used to al-
locate that virtual machine to the second computer. VIMs can also be used to move virtual
machines from one physical computer to another, in some cases even while the virtual
machine continues to operate.
Along similar lines, various types of cloud storage technologies can be used to al-
locate, move, and manage storage capacity among multiple virtual machines. Although
there are various ways of implementing this approach,20 such technologies essentially vir-
tualize storage. For example, if a user creates a “virtual drive” within a virtual machine
and writes data to it, storage technology will take that data, break it up into chunks, and
write it to multiple physical storage devices, such as hard drives, which can be on multiple
computers. Even though the chunks are spread across multiple computers and multiple
drives, the storage technology keeps track of exactly where each chunk is so that, to an
end user, those disparate chunks appear to be on one drive. In addition to keeping track
of those chunks, such technologies will also either record multiple copies of the data, or
apply an algorithm that generates additional data so that if one or more chunks are lost
(for example, due to the failure of a physical hard drive), the data can still be recovered.
Along similar lines, functions that would otherwise be served by physical networking
devices, such as firewalls, routers, and switches, can be replicated in software.21

20 Traditional approaches include, for example, storage area networks—typically proprietary hardware
consisting of one or more devices, each containing numerous drives. More recently, software solutions
have been developed to implement the same (or better) functionality using generic, nonvendor-specific
hardware, some of which is open source. These include, for example, Ceph (http://‌ceph.‌com/) and Gluster
(https://‌www.‌gluster.‌org/), both open source, as well as proprietary technologies such as VMware vSAN
(https://‌www.‌vmware.‌com/‌products/‌vsan.‌html) and a subset of the functionality in Nutanix (https://‌
www.‌nutanix.‌com/).
21 Examples of such technologies include virtual switch technology built directly within hypervisors
to enable virtual machines on a given host to communicate between each other and to the host ma-
chine and elsewhere (for example, virtual switches in VMware’s vSphere, https://‌www.‌vmware.‌com/
‌support/‌ws55/‌doc/‌ws_‌net_‌component_‌vswitch.‌html, or libvert virtual network switches in KVM,
https://‌access.‌redhat.‌com/‌documentation/‌en-‌US/‌Red_‌Hat_‌Enterprise_‌Linux/‌6/‌html/‌Virtualization_‌
Administration_‌Guide/‌chap-‌Virtualization_‌Administration_‌Guide-‌Virtual_‌Networking.‌html, as well
as software-defined networking technologies, which are typically (but not always) implemented in ded-
icated physical networking devices and shift control of certain networking functions (typically referred
to as the control plane) so that they are programmatically controlled by software and, in some cases, by
virtualizing specific network functions such as load balancers or firewalls. Example of the latter include
OpenDaylight, https://‌www.‌opendaylight.‌org/‌, Open vSwitch, http://‌openvswitch.‌org/, and VMware
NSX, https://‌www.‌vmware.‌com/‌ca/‌products/‌nsx.html.

ABA_15_Cloud_Computing_CH01.indd 6 09/04/19 3:47 PM


Overview of Cloud Computing Technology  7

Perhaps due in part to the burgeoning popularity of cloud computing, a number of

groups and companies have created collections of software tools (or “stacks”) that can be
used to comprehensively implement a cloud computing environment. These include, for
example, OpenStack,22 Apache CloudStack,23 and vCloud Suite.24

G. Recent Developments
As with most other types of information technology, cloud computing continues to devel-
op and evolve at a rapid pace. The following are some examples of recent developments
in cloud computing.
1) Cloud-native applications. The term “cloud native” is generally used to refer to ap-
plications that have been designed and written to run on cloud services, meaning
that they can run on various cloud platforms, can be easily copied or moved, and are
scalable.25 They typically will also take a “microservice” approach: instead of one large
application performing many functions, those functions are divided into numerous
individual applications (or microservices) that can interface with each other, making
each easier to maintain and allowing for a more modular approach in that end users
can combine microservices as appropriate for the task at hand.26 One of the primary
enabling technologies for cloud-native applications is container technology.27 Con-
ceptually, cloud-native applications can be thought of as abstraction at the application
level in the same way that physical machines are abstracted into virtual machines.28
2) Serverless computing. A somewhat related concept is that of “serverless comput-
ing” where the cloud provider assumes responsibility for dynamically providing
whatever resources are required by the application, rather than the end user man-
ually allocating additional resources as required. For example, with a traditional
approach, a user would create an application, select an appropriately sized virtual
machine, and run it, paying for the use of that virtual machine. If the workload
grows, then the user would increase the size of that virtual machine. In contrast, if
the same application were implemented using a serverless approach, the end user
would pay only for the resources actually used by the application. If, for example,
the application were not used for a given period of time, no resources would be
used; therefore, no usage fees would be paid.29

22 https://www.openstack.org/.
23 https://cloudstack.apache.org/.
24 https://www.vmware.com/ca/products/vcloud-suite.html.
25 Libby Clark & Mark Hinkle, Guide to the Open Cloud 6 (2016), https://www.linuxfoundation.
org/publications/guide-to-open-cloud-2016.
26 For a more detailed description of microservices architecture, see MuleSoft, Microservices vs Mono-
lithic Architecture, https://‌www.‌mulesoft.‌com/‌resources/‌api/‌microservices-‌vs-‌monolithic.
27 See Part II.D., supra.
28 Clark & Hinkle, supra note 25, at 7.
29 For a very detailed and helpful overview of serverless computing, see Mike Roberts, Serverless Archi-
tectures, Martinfowler.com, Aug. 4, 2016, https://martinfowler.com/articles/serverless.html. See also
Carl Osipov, Composable Architecture Patterns for Serverless Computing Applications—Part 4, Clouds
with  Carl,  Aug.  15,  2016,  http://‌cloudswithcarl.‌com/‌blog/‌2016/‌08/‌15/‌composable-‌architecture-‌
patterns-‌for-‌serverless-‌computing-‌applications-‌part-‌4/. Examples of serverless computing services are
AWS Lambda, https://docs.aws.amazon.com/lambda/latest/dg/welcome.html, and Google Cloud Func-
tions, https://cloud.google.com/functions/docs/.

ABA_15_Cloud_Computing_CH01.indd 7 09/04/19 3:47 PM

 8 Cloud 3.0: Drafting and Negotiating Cloud Computing Agreements

3) Edge computing. To date, most cloud computing has been enabled through the
use of large, centralized data centers hosting thousands of computers. This has
resulted in a significant number of users being rather distant from the physical
location where processing actually occurs, causing small delays in the processing
of information (sometimes referred to as latency). Although in many cases the de-
gree of latency is not critical, for certain applications, such as autonomous vehicles
and robotics, it may be. As such applications become more widely implemented,
there will be an increasing demand for low-latency cloud computing, resulting in
the increased use of decentralized data centers that are more geographically dis-
persed closer to end users and therefore lower latency.30 Interestingly, some cloud
providers have developed solutions to “extend” their cloud services to operate on
an end user’s equipment to achieve this objective.31
4) Big computing. Artificial intelligence and, in particular, deep learning typically
require significant amounts of computing power, but for limited periods of time.
As such technologies become more widely used, the demand for high-performance
computing will increase. As a result, cloud providers have begun to develop “big
compute” cloud services to enable end users to utilize deep learning without the
need to procure a high-performance data center.32
5) Decentralized cloud computing. At present, Infrastructure as a Service (IaaS)33
cloud computing is generally dominated by a small number of large companies,
each of which owns and operates multiple, large-scale data centers with millions
of computers. This is perhaps to be expected, given the advantages afforded by
economies of scale. However, there are still millions of computers and computing
devices outside of those owned by cloud providers that, for the most part, are not
used to their full capacity.34 Technologies to harness such unutilized capacity have
been used for more than two decades with a fair degree of success, but until recent-
ly have largely been limited to volunteer efforts for scientific research.35 ­However, a

30 Jeremy Hsu, It’s Time to Think Beyond Cloud Computing, Wired, Aug. 23, 2017, https://www.wired.
com/story/its-time-to-think-beyond-cloud-computing.
31 For example, AWS Greengrass, https://aws.amazon.com/greengrass, and Microsoft Azure IoT Edge,
https://azure.microsoft.com/en-ca/services/iot-edge/.
32 Amazon and Microsoft both now offer high-performance cloud computing. See High Perfor-
mance Computing, Amazon, https://‌aws.‌amazon.‌com/‌hpc/; High Performance Computing, Microsoft,
https://‌azure.‌microsoft.‌com/‌en-‌us/‌solutions/‌high-‌performance-‌computing/.
33 For an explanation of Infrastructure as a Service, see Part III.B., infra.
34 Smartphones alone constitute a significant portion of this capacity. In 2013, one researcher stated that
“there are about a billion Android devices right now, and their total computing power exceeds that of the
largest supercomputers.” See Richard Gray, How CERN’s Grid May Place the Power of the World’s Computers
in Your Hands, Telegraph, Aug. 14, 2013, http://‌www.‌telegraph.‌co.‌uk/‌technology/‌news/‌10242837/‌How-
‌CERNs-‌Grid-‌may-‌place-‌the-‌power-‌of-‌the-‌worlds-‌computers-‌in-‌your-‌hands.‌html. As of 2018, there are
approximately 2.5 billion active smartphones. See Number of Smartphone Users Worldwide from 2014 to
2020, Statista, https://‌www.‌statista.‌com/‌statistics/‌330695/‌number-‌of-‌smartphone-‌users-‌worldwide/.
35 The first such effort was the Great Internet Mersenne Prime Search, https://‌www.‌mersenne.‌org/,
launched in 1996 to (as the name suggests) search for Mersenne prime numbers, which takes a significant
amount of computing power. The project has discovered 15 prime numbers to date. In 2017, slightly over
1.5 million CPUs participated in the project. Other similar projects include Distributed.net, http://‌
www.‌distributed.‌net/‌Main_‌Page, which undertakes research projects primarily related to cryptography;
SETI@home, https://‌setiathome.‌berkeley.‌edu/, which uses computing power to analyze radio telescope
data for signs of extraterrestrial intelligence; and Folding@home, http://folding.‌stanford.‌edu/, which fo-
cuses on disease research by simulating the folding of proteins and other elements.

ABA_15_Cloud_Computing_CH01.indd 8 09/04/19 3:47 PM


Overview of Cloud Computing Technology  9

number of ventures have recently been launched to offer general IaaS cloud com-
puting services powered by decentralized computing resources, using blockchain
technology to assist in accounting for the usage and provisioning of resources.36
Such an approach has some possible advantages as compared with traditional
cloud computing, such as lower cost and higher responsiveness,37 although it may
too early to assess whether those advantages will be realized at any significant scale.

III. Characteristics and Models

While Part II above focused on the elements of technology that are used to create cloud
computing solutions, Part III outlines the general characteristics of cloud computing from
the perspective of the user, and describes how cloud computing services are characterized
both in terms of the nature of the services being provided (the “service model”) and who
is entitled to use such services (the “deployment model”).

A. Characteristics of Cloud Computing Services

Application of the technologies described above results in the availability of computing
services that have a specific set of characteristics. The NIST Definition identifies five such
characteristics:38
1) On-demand self-service. End users can easily obtain the computer services
through automated means (such as a website) and without human assistance.
2) Broad network access. Services are provided through standardized means (for
example, the internet and web browsers) and can be accessed and used utilizing a
variety of devices, such as mobile phones, tablets, or desktop computers.
3) Resource-pooling. Computing resources are pooled together and dynamically
assigned (and reassigned) for use by multiple users, generally without knowledge
or control on the part of the user regarding where those resources are located.
4) Rapid elasticity. The quantity of resources can easily and quickly (and in some cases
automatically) increase or decrease on demand in almost any quantity and at any time.
5) Measured service. Billing is metered and delivered as a utility service.

B. Service Models
The NIST Definition also sets out three service models under which cloud services are
categorized based on the nature of the services provided:39

36 Examples include Blockchain-Based Decentralized Cloud Computing, iExec, http://iex.ec/, and

Dfinity, https://‌dfinity.‌network/, both of which offer (or plan to offer) general IaaS services. Storj,
https://‌storj.‌io/, focuses on distributed cloud storage.
37 For further details on how such advantages could be realized, see Storj Labs, THERE IS NO CLOUD:
It’s Just Someone Else’s Computer, Medium, Mar. 16, 2016, https://‌medium.‌com/@‌storjproject/‌there-‌is-‌
no-‌cloud-‌it-‌s-‌just-‌someone-‌else-‌s-‌computer-‌6ecc37cdcfe5; and Gilles Fedak, How Can Blockchain
Improve Cloud Computing, Medium, Sept. 1, 2016, https://‌medium.‌com/‌iex-‌ec/‌how-‌blockchain-‌can-‌
improve-‌cloud-‌computing-‌1ca24c270f4f.
38 Paraphrased from Mell & Grance, supra note 6, at 2.
39 Paraphrased from id. at 2–3.

ABA_15_Cloud_Computing_CH01.indd 9 09/04/19 3:47 PM

 10 Cloud 3.0: Drafting and Negotiating Cloud Computing Agreements

1) Software as a Service (SaaS). The provision of application services running on

a cloud infrastructure,40 typically accessed by end users through a web browser,
without any ability on the part of the end user to manage or control the infra-
structure used to provide such services. Examples include e-mail services (such
as Google Gmail41) and customer relationship management applications (such
as Salesforce42). SaaS generally comprises the largest proportion of spending on
cloud computing.43
2) Platform as a Service (PaaS). The provision of programming languages and tools
to enable end users to create applications and deploy those applications on a cloud
platform for use by the user or by others. As with SaaS, the end user does not
manage or control the underlying infrastructure. Examples include Heroku44 and
Google App Engine.45
3) Infrastructure as a Service (IaaS). The provision of fundamental computing re-
sources such as processing, storage, and networking to enable users to run any
software they choose, including operating systems or applications, but with no
control or management of the underlying infrastructure other than the specific
computing resources provisioned. Examples include Amazon EC246 (computing),
Amazon S347 (storage), and Microsoft Azure.48
A given cloud service provider does not necessarily own or control all the resources
used to provide its services. For example, a SaaS vendor may decide to implement its ap-
plication using the services provided by a separate IaaS vendor so that the SaaS vendor can
focus on its strengths (namely, the development of the application) rather than non-core
competencies, such as building and managing data centers and physical servers.49 That
IaaS vendor may, in turn, decide to use a third-party vendor to provide data centers and
hardware.50

40 The NIST Definition defines a cloud infrastructure as one that meets the five characteristics described
in Part III.A., supra.
41 https://www.google.com/gmail/about/.
42 https://‌www.‌salesforce.‌com/.
43 SaaS represented approximately 73 percent of worldwide spending on public cloud computing in
2015. IaaS represented 16 percent, and PaaS represented 11 percent. See Gantz & Miller, supra note 2,
at 4.
44 https://‌www.‌heroku.‌com/.
45 https://‌cloud.‌google.‌com/‌appengine/.
46 https://aws.‌amazon.‌com/‌ec2/.
47 https://‌aws.‌amazon.‌com/‌s3/.
48 https://azure.microsoft.com/. Both Amazon AWS and Microsoft Azure include many different offer-
ings, some of which are IaaS in nature, whereas others are more properly described as PaaS or SaaS.
49 See, e.g., Barb Darrow, Welcome to the Era of Great Data Center Consolidation, Fortune, Feb. 15,
2017, http://‌fortune.‌com/‌2017/‌02/‌15/‌data-‌center-‌consolidation-‌cloud/. The article explains that soft-
ware developers are increasingly relying on IaaS vendors to deliver the functionality of their software. Jeff
Bezos, CEO of Amazon, has compared this to the evolution of the power industry: “You go back in time
a hundred years, if you wanted to have electricity, you had to build your own little electric power plant,
and a lot of factories did this. As soon as the electric power grid came online, they dumped their electric
power generator, and they started buying power off the grid. It just makes more sense. And that’s what is
starting to happen with infrastructure computing.” Brad Stone, The Everything Store: Jeff Bezos
and the Age of Amazon (2013).
50 Most IaaS vendors, however, own and control their own physical infrastructure.

ABA_15_Cloud_Computing_CH01.indd 10 09/04/19 3:47 PM


Overview of Cloud Computing Technology  11

C. Deployment Models
The NIST Definition also sets out four different deployment models that describe who is
permitted to use the infrastructure powering the cloud service:51
1) Private cloud. The infrastructure can be used by only a single organization (which
may be comprised of multiple end users). It may be owned, managed, or operated
by the organization itself, a third party, or a combination thereof, and the physical
infrastructure may be located on the premises of that organization or elsewhere.
Private clouds are typically implemented where isolation of the physical resourc-
es is required—for example, to facilitate higher security or for regulatory com-
pliance.52 The trade-off is that private clouds are more costly and less flexible as
compared to public clouds.53 The end user typically will pay for all the dedicated
physical resources comprising the private cloud and, if the capacity of the private
cloud is reached, must pay for additional physical resources to be installed and
added to the cloud.54
2) Community cloud. Similar to private cloud except that use is limited to a specific
community whose members have shared concerns, interests, or needs.
3) Public cloud. A cloud service that is available for use by the general public.55 The
infrastructure may be owned, managed, and operated by a business, governmen-
tal, or other organization (or a combination thereof). The physical infrastructure
is typically located on the premises of the provider.
4) Hybrid cloud. A combination of two or more of the above, with technology that
enables data and applications to be moved or shared between or among them.
Although the NIST Definition only contemplates the use of this term to refer-
ence combinations of different cloud types, the term “hybrid” also is often used
to describe offerings that enable on-premise computing resources to access cloud
services, such as additional storage. End users can install either an appliance or
software on premises, which will then enable them to access cloud storage as if it
were an on-site physical storage device.56

51 Paraphrased from Mell & Grance, supra note 6, at 3.

52 Some vendors use the term “private cloud” to mean something different. In some cases, offerings
marketed as private clouds do, in fact, involve shared resources that are logically (but not physically) seg-
regated. Care should be taken in investigating offerings described as private clouds to ensure the degree
and nature of isolation meets the requirements of the end user.
53 For a useful summary of trade-offs between private and public clouds, see Comparing Private and
Public Clouds, Rackspace, https://www.rackspace.com/library/cloud-computing-difference.
54 Some vendors have begun to offer utility pricing (i.e., pricing based on use) for private clouds—see,
e.g., Arturo Suarez, Introducing Utility Pricing to Your Managed Private Cloud, Ubuntu, July 9, 2015,
https://‌insights.‌ubuntu.‌com/‌2015/‌07/‌09/‌introducing-‌utility-‌pricing-‌to-‌your-‌managed-‌private-‌cloud/,
but such pricing is only for management and operation services and does not include the underlying
hardware, which the end user must purchase separately.
55 Availability to the general public does not necessarily mean that the public cloud services purchased
by a given user are made available to the general public, or that they are made freely available to the gen-
eral public.
56 Examples include Microsoft Azure StorSimple, https://‌azure.‌microsoft.‌com/‌en-‌us/‌services/
‌storsimple/, and Amazon AWS Storage Gateway, https://‌docs.‌aws.‌amazon.‌com/‌storagegateway/
‌latest/‌userguide/‌WhatIsStorageGateway.‌html.

ABA_15_Cloud_Computing_CH01.indd 11 09/04/19 3:47 PM

 12 Cloud 3.0: Drafting and Negotiating Cloud Computing Agreements

Some cloud providers have developed slight variations on the above models. One
example is a virtual private cloud, a pool of computing resources created within a public
cloud infrastructure but subject to a higher degree of isolation than usual to ensure the
pool is available to only the user creating the virtual private cloud.57
Virtual private clouds should not be confused with virtual private servers. Despite the
similar name and the use of some of the same underlying technology, virtual private serv-
ers are distinct from virtual private clouds, or cloud computing generally for that matter.
A virtual private server is essentially a virtual machine created and hosted on a computer
that is allocated for the exclusive use of a given user. In other words, the end user is pur-
chasing a portion of the resources of a physical computer that has been divided up into
multiple virtual machines. In contrast to cloud computing, there are typically limits to the
scalability of a given virtual private server, and if there is a failure of the host computer, all
virtual private servers on that computer will also fail.
Another variation of the above models is a multicloud approach, where the services of
two or more cloud service providers are used in tandem. For example, an end user could
configure a primary virtual server with one service provider and a backup server with
another, and then configure them to allow the backup server to automatically take over
in the event the primary virtual server fails. Some organizations have developed solutions
that facilitate the implementation of and help manage multicloud approaches.58
In addition, some organizations have begun working toward the development of
cloud interoperability—an approach sometimes referred to as “intercloud.” In contrast
to a multi-cloud approach, where the end user or its service provider assumes the re-
sponsibility for developing solutions to address the specific standards of each different
cloud service vendor (which may change over time), an intercloud approach focuses on
the development of standardized interfaces to be adopted by cloud service providers to
enable interoperability among themselves, and to allow users to easily access multiple
cloud service providers though that same interface.59 To this end, the Institute of Electrical
and Electronics Engineers has begun developing standards to help facilitate the adoption
of intercloud computing.60 However, to date, intercloud does not appear to have gained
much traction with cloud providers.61

57 This is typically accomplished through the use of a separate virtual network along with additional
security measures. See, e.g., Amazon Virtual Private Cloud, Amazon, https://‌aws.‌amazon.‌com/‌vpc/‌.
58 See, e.g., Cloud Foundry, https://‌www.‌cloudfoundry.‌org/‌multi-‌cloud/‌; interCloud, https://‌www.‌inter-
cloud.‌com/; and RightScale, https://‌www.‌rightscale.‌com/. See also James Sanders, Multicloud: A Cheat Sheet,
TechRepublic, May 4, 2017, http://‌www.‌techrepublic.‌com/‌article/‌multicloud-‌the-‌smart-‌persons-‌guide/.
59 Cloud Strategy Partners, LLC, IEEE Intercloud Interoperability and Federation Frame-
work   6–8,   https://‌cloudcomputing.‌ieee.‌org/‌images/‌f iles/‌education/‌studygroup/‌IEEE_‌Intercloud_‌
Interoperability_‌and_‌Federation_‌Framework.‌pdf (suggesting that cloud computing interoperability is
inevitable, following similar trends toward uniform standards in various other industries, such as electric
power, financial markets, telephony, and the internet).
60 See IEEE Project P2302—Standard for Intercloud Interoperability and Federation, IEEE, https://
‌standards.‌ieee.‌org/‌develop/‌project/‌2302.‌html; IEEE Project P2301—Guide for Cloud Portability and
Interoperability Profiles, IEEE, http://‌standards.‌ieee.‌org/‌develop/‌project/‌2301.‌html.
61 For example, Cisco shut down its intercloud offering in early 2017. Simon Sharwood, Cisco to Kill Its Inter-
cloud Public Cloud on March 31, 2017, Register, Dec. 13, 2016, https://‌www.‌theregister.‌co.‌uk/‌2016/‌12/‌13/
‌cisco_‌to_‌kill_‌its_‌intercloud_‌public_‌cloud_‌on_‌march_‌31st_‌2017/. Although one possible advantage of
intercloud would be to enable smaller cloud providers to present themselves collectively as a much larger
network, it is less clear why market-dominant cloud providers would be interested in intercloud. Some
vendors, however, do continue to market intercloud connectivity. See, e.g., Multi-Cloud Connectivity with
RackConnect Global, RackConnect, https://‌www.‌rackspace.‌com/‌hybrid/‌rackconnect/‌global.

ABA_15_Cloud_Computing_CH01.indd 12 09/04/19 3:47 PM


Overview of Cloud Computing Technology  13

IV. Advantages of Cloud Computing

Although the technology that enables cloud computing is critical, business considerations
are of equal importance. Innovative technology is unlikely to be adopted if doing so does
not make business sense. This section outlines some of the business advantages of cloud
computing.

A. Costs and Cash Flow

Many cloud providers, particularly those offering IaaS, price their services on a utility
basis: there are no up-front costs, and fees are charged only as and when the services are
used.62 This enables end users to more precisely align their expenses with their require-
ments.63 For example, a small business can begin with a small virtual machine provisioned
at a low cost. As its business grows, it can increase the size of the virtual machine in small
increments or purchase additional virtual machines fairly quickly and easily. If there is a
downturn in business, the company can also quickly scale down its usage, and therefore
its costs, accordingly. If there is a significant project that requires an unusually high level
of computing resources, it can temporarily procure as much as it may require through the
cloud provider. Given that costs incurred are based on the needs of the business at a given
point in time, expenditures are more likely to align with revenue.
Compare this to the more traditional model in which an organization must purchase
a server and license software at a significant up-front cost. Given that the quantity of
resources purchased must be sufficient for peak usage, and given that those resources
(depending on the business) may not be used on a 24/7 basis, the resources will typically
be underutilized. As its business grows, it must buy a minimum of one additional server,
even if it does not require all of the capacity of an additional server. It must also factor
in room for growth because the process of purchasing, installing, and configuring an ad-
ditional server takes substantial time. If there is a downturn in business, the company is
unlikely to realize much value when it disposes of unneeded servers. Given that the ini-
tial capital outlays do not necessarily align with revenue, financing may also be required.
On-premise hardware will also require other costs, such as supporting hardware (for ex-
ample, uninterruptible power supplies), space, staff, electricity, and possibly support and
maintenance contracts requiring long-term financial commitments.64

B. Efficiency—Economies of Scale and Commodification

Of course, hardware must still be purchased to enable cloud computing, and although
end users who purchase cloud services are relieved of those capital expenditures, cloud

62 This is less likely to be the case for SaaS providers. Although some SaaS providers will allow for a
degree of flexibility in increasing or decreasing usage, there will typically be some limitations, such as a
minimum subscription period or minimum usage requirements, and the offering will therefore not be as
flexible as most public cloud IaaS offerings.
63 This may not be true for private cloud deployments. See Part III.C., supra.
64 For a more in-depth economic analysis of cloud versus on-premise equipment, see Cloud Economics—
Are You Getting the Bigger Picture?, Cloud Technology Partners, https://‌www.‌cloudtp.‌com/‌doppler/
‌cloud-‌economics-‌getting-‌bigger-‌picture/.

ABA_15_Cloud_Computing_CH01.indd 13 09/04/19 3:47 PM

 14 Cloud 3.0: Drafting and Negotiating Cloud Computing Agreements

providers (and in particular IaaS providers)65 are not. Even though they face the same
challenges as end users regarding capital expenditures, cloud providers have been able to
maintain competitive pricing by becoming more efficient in their delivery of services in
a number of ways:
1) Hardware Costs. Cloud providers can rely on economies of scale to realize sav-
ings due to the volume of their purchases. In addition, rather than purchasing
the brand-name, premium-priced equipment that enterprises typically purchase,
cloud providers purchase lower-priced commodity hardware because resiliency
and redundancy is built into the enabling technology they use. Some cloud pro-
viders have developed their own custom hardware designs to maximize the cost
efficiency of their hardware purchases.66
2) Software Costs. Many IaaS cloud providers make extensive use of open source
software, thus eliminating (or greatly reducing) the cost of software.67 In addition,
certain information technology functions that were traditionally performed by
hardware are now handled through the use of software.68
3) Efficient Resource Usage. An inherent function of the technology used to enable
the provision of cloud computing is efficient resource allocation and utilization
across physical resources.69 IaaS providers can also balance peaks and valleys in
resource usage over time and geography across their customer base. The net result
is that IaaS providers can operate much more efficiently than enterprises do—by
some estimates up to three times more efficiently.70

C. Elasticity and Scalability

IaaS users can provision one or many virtual server in minutes and can add or remove
resources within similar timeframes. In many cases, resources can be configured to

65 SaaS providers can, of course, run their applications using an IaaS provider rather than building out
their own physical infrastructure—and many do. See Julie Bort, Netflix, Juniper, and Intuit Explain How
Amazon Is Eating the $3.5 Trillion IT Industry, Business Insider, Jan. 13, 2016, http://‌www.‌businessinsider.‌
com/‌netflix-‌intuit-‌juniper-‌go-‌all-‌in-‌on-‌amazon-‌cloud-‌2016-‌1.
66 See, e.g., Dan Richman, Amazon Web Services’ Secret Weapon: Its Custom-Made Hardware and Net-
work, GeekWire, Jan. 19, 2017, https://‌www.‌geekwire.‌com/‌2017/‌amazon-‌web-‌services-‌secret-‌weapon-
‌custom-‌made-‌hardware-‌network/.
67 For example, Amazon Web Services appears to make extensive use of Linux. Steven J. Vaughan-
Nichols, Amazon EC2 Cloud Is Made Up of Almost Half-a-Million Linux Servers, ZDNet, Mar. 16, 2012,
http://‌www.‌zdnet.‌com/‌article/‌amazon-‌ec2-‌cloud-‌is-‌made-‌up-‌of-‌almost-‌half-‌a-‌million-‌linux-‌servers/.
More than 20 public cloud providers use OpenStack, an open-source cloud computing platform. Barb
Darrow, HPE and Cisco Moves Hurt OpenStack’s Public Cloud Story, Fortune, Dec. 19, 2016, http://
‌fortune.‌com/‌2016/‌12/‌19/‌openstack-‌public-‌cloud/. Google uses open-source code such as Ganeti,
https://‌opensource.‌google.‌com/‌projects/‌ganeti, and Kubernetes, https://kubernetes.io/, both origi-
nally developed by Google, although the extent to which they are used in its cloud computing offer-
ings is unclear. Google has, however, indicated that it uses Linux extensively in its cloud offerings. See
Cade Metz, Google Shaman Explains Mysteries of “Compute Engine,” Wired, July 3, 2012, https://‌www.
‌wired.‌com/‌2012/‌07/‌google-‌compute-‌engine/.
68 See Part II.F., supra.
69 See Part II.D., supra.
70 See Larry Dignan, AWS Cloud Computing Ops, Data Centers, 1.3 Million Servers Creating Efficiency
Flywheel,  ZDNet,   June   7,   2016,   http://‌www.‌zdnet.‌com/‌article/‌aws-‌cloud-‌computing-‌ops-‌data-‌centers-‌
1-‌3-‌million-‌servers-‌creating-‌efficiency-‌flywheel/.

ABA_15_Cloud_Computing_CH01.indd 14 09/04/19 3:47 PM


Overview of Cloud Computing Technology  15

automatically scale based on workloads. Compare this to a typical process to procure

a physical server, which would likely involve several days at a minimum. Unlike physi-
cal hardware, there is no need to install or remove servers, nor are there concerns about
hardware failures. In brief, cloud computing is more flexible and convenient for end users.

D. Increased Functionality and Reliability

One aspect of cloud computing that has often been cited by Amazon in relation to its
cloud offerings is what it has dubbed the “flywheel effect” or the “virtuous circle.” Cloud
computing features significant economies of scale. As a cloud provider signs on more
customers, there is more usage; more usage leads the cloud provider to roll out more
infrastructure; more infrastructure leads to greater economies of scale, resulting in lower
infrastructure costs; lower infrastructure costs lead to additional innovations and reduced
prices, both of which in turn lead to more customers, and so on. The net result for Ama-
zon has been an exponential increase in new features over time—for example, it launched
a total of 24 new services in 2008, but in 2015 it launched 516.71 Amazon’s early emphasis
on developing its cloud computing business has given it a substantial lead in this efficien-
cy race, which some believe is unlikely to be matched by any other provider.72
The adoption of a SaaS as a delivery mechanism, as compared to more traditional
software licensed for on-premise use,73 has given SaaS providers unprecedented insight
into the usage of their applications. SaaS vendors can gather great quantities of data on
how their applications are used, who is using them, when they are used, what features are
used (and which are not), problems encountered, and so on, all in real time.74 As a result,
SaaS providers are able to identify and fix bugs sooner and identify and improve the fea-
tures and functionality of their application more rapidly. In addition, SaaS users have few-
er concerns about patches, version management, and compatibility and interoperability
with hardware, operating systems, and other software.

E. Additional Revenue Streams

Cloud computing approaches also open up the possibility of realizing additional revenue
streams for providers, particularly for SaaS providers. Although this is of course primarily
of benefit to providers, it may indirectly benefit users by enabling the cloud provider to
lower fees charged to end users for the cloud provider’s services. The following are a few
examples:

71 Matt Weinberger, The Cloud Wars Explained: Why Nobody Can Catch Up with Amazon, Business In-
sider, Nov. 7, 2015, http://‌www.‌businessinsider.‌com/‌why-‌amazon-‌is-‌so-‌hard-‌to-‌topple-‌in-‌the-‌cloud-
‌and-‌where-‌everybody-‌else-‌falls-‌2015‌-10.
72 Id.
73 Collection of usage information is also possible with on-premise software (often referred to as “te-
lemetry”) and has become more prevalent over time. However, it is subject to certain challenges, e.g., the
blocking of telemetry transmissions by an end user’s security measures, such as firewalls.
74 The rise of analytics has resulted in the growth of service providers that enable SaaS vendors to im-
plement monitoring more easily. Examples include AppDynamics, https://‌www.‌appdynamics.‌com/‌info/‌
service-‌saas/, KISSmetrics, ‌https://‌blog.‌kissmetrics.‌com/‌saas-‌guides/, and Mixpanel, https://‌mixpanel.
‌com/‌solutions/‌saas/.

ABA_15_Cloud_Computing_CH01.indd 15 09/04/19 3:47 PM

 16 Cloud 3.0: Drafting and Negotiating Cloud Computing Agreements

1) Data monetization. In contrast to on-premise software, with SaaS the service

provider will typically have some degree of access to the data of its users. This
opens up the possibility of utilizing the data for purposes other than the provision
of services to the end user in order to generate additional revenue. Although the
topic is not without controversy and often raises the hackles of prospective SaaS
users, data monetization is not a novel concept and is a strategy that has been
employed by other types of service providers since before the advent of cloud
computing.75 In some cases, the potential value in data monetization could exceed
the revenue generated from the service itself.76 It is not difficult to envision how
this could be applied to various types of SaaS applications across numerous indus-
tries.77 Of course, end users may have concerns regarding privacy, confidentiality,
and security of their data, particularly in the case of highly sensitive data such as
personal health information.
2) Networks, ecosystems, and marketplaces. Although not necessarily exclusive to
cloud offerings, cloud providers are to some extent better positioned to enable in-
tegration with third-party software or solutions, due in part to the use of industry
standard interfaces and protocols.78 Cloud providers can further encourage this
by developing interfaces and protocols to enable third parties to easily integrate
with the cloud provider’s offerings, or offer PaaS solutions to enable the develop-
ment or migration of their applications.79 Doing so allows the cloud provider to
earn additional revenue by charging for the privilege of integrating with the cloud
provider’s platform, for example, or acting as a reseller or distributor and earning

75 For example, ADP, a payroll service provider, leverages the information it collects in the course of pro-
viding such services to develop analytics and statistical reports and related services. See Tom Davenport,
How ADP Gives Data Value Back To Its Customers, Forbes, Mar. 12, 2018, https://www.forbes.com/sites/
tomdavenport/2018/03/12/how-adp-gives-data-value-back-to-its-customers/#2d4830a5732c. American
Express has used transaction information from its 90 million cards to create Business Insights, “a global
information, analytics and consulting organisation that combines real behavioural information—based
on actual aggregated purchasing data—with sophisticated analytics to reveal unique insights about your
customers, competitive set and marketplace.” Attract Customers/Business Insights, American Express,
https://‌www.‌americanexpress.‌com/‌uk/‌content/‌merchant/‌business-‌insights.‌html. Financial institutions
and credit bureaus have a long history of sharing and aggregating data to produce credit reports.
76 One possible example, albeit not from the SaaS world, is Nest, a maker of smart thermostats. Nest
collects and aggregates information on energy consumption from those thermostats to provide aggre-
gated energy consumption data to energy companies to enable them to better plan their energy produc-
tion. The founder of Nest believes that the sale of such data will eventually earn more than the sales of
the thermostats. See Matthew Mobrea, Google’s Real Plan Behind the Purchase of the Nest Thermostat,
ITworld, Apr. 25, 2014, https://‌www.‌itworld.‌com/‌article/‌2833423/‌big-‌data/‌google-‌s-‌real-‌plan-‌behind-
‌the-‌purchase-‌of-‌the-‌nest-‌thermostat.‌html.
77 For example, an online sales or e-commerce platform (e.g., Shopify) could develop analytics on retail
sales and pricing strategy; an online accounting services provider could develop analytics on financial
information and lending. See also Leo Polovets, The Value of Data, Part 3: Data Business Models, Coding
VC, Mar. 12, 2015, https://‌codingvc.‌com/‌the-‌value-‌of-‌data-‌part-‌3-‌data-‌business-‌models. The article
discusses other companies that currently use such strategies.
78 See Part II.D., supra. See also Rossey Charleston, Cloud vs On-Premise Software Integrations,
B2B Commerce Digital, Nov. 6, 2015, https://www.handshake.com/blog/cloud-vs-on-premise-
integrations/.
79 See, e.g., AWS SaaS Partner Program, Amazon, https://‌aws.‌amazon.‌com/‌partners/‌saas/.

ABA_15_Cloud_Computing_CH01.indd 16 09/04/19 3:47 PM


Overview of Cloud Computing Technology  17

a margin when its users purchase third-party solutions.80 The availability of an

ecosystem comprised of the cloud provider’s offerings integrated with third-party
solutions may be attractive to end users insofar as it offers one-stop shopping and
may reduce the efforts otherwise needed to integrate disparate solutions. This, of
course, may make it difficult for end users to transition to another cloud provid-
er—a benefit for the cloud provider, but perhaps a detriment to the end user.
3) Advertising. Cloud providers can also potentially earn revenue through the place-
ment of advertisements within their offerings, but at the risk of alienating users.
In contrast to passive, content-only websites, SaaS providers may be in a position
to leverage the context of the offered services or end user data to target ad place-
ments much more precisely. For example, users of an online accounting solution
could be presented with advertisements for loans or tax-return preparation.81

V. Disadvantages of Cloud Computing

Cloud computing entails certain disadvantages and risks. The following are some of the
more typical concerns:
1) Reliance on internet connectivity. Given that cloud services are almost always
accessed through the internet, continued access will be dependent on the reliabil-
ity of both the internet and the end user’s internet service provider. If either fails,
cloud services cannot be used. Although the risk of a failure of a given internet
service provider may be mitigated by using multiple service providers, doing so
would come at an additional cost and does not address widespread internet out-
ages. Lastly, given the nature of cloud services, additional bandwidth will likely be
required, which will increase costs.
2) Privacy and security. The use of cloud services will in almost all cases result in
the transfer of data to the cloud service provider, which may raise concerns about
the cloud provider’s ability to protect against data breach. Cloud providers will
typically implement a standard set of security measures, which may or may not
meet the requirements of a given user. Although larger providers may implement
a fairly robust set of security measures,82 users must nonetheless evaluate the extent

80 One example of this is Amazon’s AWS Marketplace, https://‌aws.‌amazon.‌com/‌marketplace/, which

offers thousands of software options that can be deployed on its cloud platform, including operating
systems, security software, databases, and business intelligence applications. At the SaaS level, Salesforce’s
AppExchange, https://‌appexchange.‌salesforce.‌com, offers thousands of applications and services that can
be used seamlessly within its platform. Note that the possibilities are not limited to software applica-
tions. For example, a SaaS provider offering supply-chain management applications could integrate with
providers of shipping or warehousing services. Nor are they limited to providers. For example, a SaaS
provider could offer sales solutions for insurance agents as well as services for insurance companies, both
of whom would benefit from the interoperability of a common platform.
81 Perhaps one of the earliest examples of this was introduction of advertisements in Google’s Gmail.
Google displayed targeted advertising by scanning e-mail contents, although only for users of the free
consumer version. However, this practice was not without controversy, and ultimately Google decided to
discontinue the practice. See Daisuke Wakabayashi, Google Will No Longer Scan Gmail for Ad Targeting,
N.Y. Times, June 23, 2017, https://‌www.‌nytimes.‌com/‌2017/‌06/‌23/‌technology/‌gmail-‌ads.‌html.
82 See, e.g., the rather lengthy lists of privacy and security certifications for Amazon Web Services,
https://‌aws.‌amazon.‌com/‌compliance/,   Microsoft   Azure,   https://‌azure.‌microsoft.‌com/‌en-‌ca/‌support/‌
trust-‌center/, and Salesforce, https://‌trust.‌salesforce.‌com/‌en/‌compliance/.

ABA_15_Cloud_Computing_CH01.indd 17 09/04/19 3:47 PM

 18 Cloud 3.0: Drafting and Negotiating Cloud Computing Agreements

to which such measures are adequate for their specific needs, and the extent
to which providers contractually commit to responsibility for such measures. The
question of security should also be considered in comparison to what security
measures a given end user would take in the alternative—for example, if an end
user were to set up its own on- premises equipment instead of using a cloud pro-
vider, would it implement security measures that are more robust than those im-
plemented by the cloud provider?83 Lastly, end users may wish to consider adding
their own security measures to those offered by the cloud service provider, such
as encrypting data before transmitting it to the cloud.84
3) Reliance on the cloud provider. The use of a cloud provider will necessarily re-
sult in some degree of reliance on the provider. If there is a failure in the cloud
provider’s service, an end user’s business operations will likely be impacted. This
can be mitigated to some extent by using multiple redundant service providers,
though typically only for IaaS and at additional cost and complexity. However,
as with security considerations, this question should also be considered in com-
parison with the alternatives. For example, if an end user were to set up its own
on-premise equipment instead of using a cloud provider, it would still be reliant
on the vendor of that equipment to address a failure.

VI. Conclusion
This chapter is an admittedly cursory overview of the more significant technologies used
in cloud computing. It is by no means comprehensive, but will equip business lawyers with
a basic technological understanding of cloud computing to counsel more effectively when
advising on cloud computing transactions.

83 Some have argued that, in general terms, cloud providers will be better positioned to implement
robust security than their end users due, in part, to economies of scale. For example, cloud providers can
design and build physically secure data centers and employ teams of security experts and spread that cost
across their user base; given that they can monitor a much greater amount of internet traffic, they may
be able to detect malware and similar threats sooner, and the larger scale of a cloud provider’s infrastruc-
ture may be less susceptible to distributed denial of service attacks. See Andrew Froehlich, Why Cloud
Security Beats Your Data Center, Information Week, July 21, 2015, https://‌www.‌informationweek.‌com/
‌cloud/‌infrastructure-‌as-‌a-‌service/‌why-‌cloud-‌security-‌beats-‌your-‌data-‌center/‌d/‌d-‌id/‌1321354; Brandon
Butler, Public Cloud vs. On-Premises, Which Is More Secure?, NetworkWorld, Dec. 17, 2015, https://
‌www.‌networkworld.‌com/‌article/‌3016673/‌public-‌cloud/‌public-‌cloud-‌vs-‌on-‌premises-‌which-‌is-‌more-
secure.html. It is notable that U.S. intelligence agencies concluded that Amazon Web Services could
maintain adequate security when they awarded it a contract for the provision of cloud services, see Frank
Konkel, The Details About the CIA’s Deal With Amazon, Atlantic, July 17, 2014, https://‌www.‌theatlantic.
com/‌technology/‌archive/‌2014/‌07/‌the-‌details-‌about-‌the-‌cias-‌deal-‌with-‌amazon/‌374632/, although that,
of course, is not necessarily indicative of the security measures that all end users can expect. Lastly, many
recent data breaches that involve cloud services resulted from end user misuse or misconfiguration. See,
e.g., Dan O’Sullivan, The RNC Files: Inside the Largest US Voter Data Leak, UpGuard, Dec. 20, 2017,
https://‌www.‌upguard.‌com/‌breaches/‌the-‌rnc-‌files‌.
84 See Stephen Lawton, Cloud Encryption: Using Data Encryption in the Cloud, Tom’s IT Pro, Apr. 30,
2015, http://‌www.‌tomsitpro.‌com/‌articles/‌cloud-‌data-‌encryption,‌2-‌913.‌html. User-controlled security
measures may be more easily implemented for IaaS than for SaaS because end users will have a greater
degree of control over infrastructure for the former. Some vendors have developed resources that enable
end users to control encryption in connection with SaaS services. See, e.g., Vaultive’s solution for Micro-
soft Office 365, https://‌vaultive.‌com/‌for-‌your-‌technology/‌office-‌365-‌security/.

ABA_15_Cloud_Computing_CH01.indd 18 09/04/19 3:47 PM