57
Chapter 3
Building Cloud Networks
3.1 Chapter Overview
In previous chapters we have explained what cloud computing is. In this
chapter, we will describe what it takes to build a cloud network. You will
learn how and why companies build these highly automated private cloud
networks providing resources that can be managed from a single point. We
will discuss the significant reliance of cloud computing architectures on
server and storage virtualization as a layer between applications and distrib-
uted computing resources. You will learn the basics of how flexible cloud
computing networks such as those modeled after public providers such as
Google and Amazon are built, and how they interconnect with corporate IT
private clouds designed as service-oriented architectures (SOAs). We pro-
vide an overview of how SOA is used as an intermediary step for cloud com-
puting and the basic approach to SOA as it applies to data center design. We
then describe the role and use of open source software in data centers. The
use and importance of collaboration technologies in cloud computing archi-
tectures is also discussed. Last and most important, you will gain an under-
standing of how the engine of cloud computing will drive the future of
infrastructure and operations design.
Ten years ago, no one could have predicted that the cloud (both hard-
ware and software) would become the next big thing in the computing
world. IT automation has evolved out of business needs expressed by cus-
tomers to infrastructure management and administrators. There has never
been a grand unified plan to automate the IT industry. Each provider,
responding to the needs of individual customers, has been busily building
technology solutions to handle repetitive tasks, respond to events, and pro-
duce predictable outcomes given certain conditions. All the while this evo-
lutionary process was occurring, it was presumed that the cost of not doing
it would be higher than just getting it done.
1
The solutions provided to
Chap3.fm Page 57 Friday, May 22, 2009 11:25 AM
58 Cloud Computing
meet customer needs involved both hardware and software innovation and,
as those solutions emerged, they gave rise to another generation of innova-
tion, improving on the foundation before it. Thus the effects of Moore’s
law
2
seem to prevail even for cloud evolution.
From the military use of TCP/IP in the 1960s and 1970s to the devel-
opment and emergence of the browser on the Internet in the late 1980s and
early 1990s, we have witnessed growth at a rate similar to what Gordon
Moore had predicted in 1965: essentially, a doubling of capability approxi-
mately every two years. We saw the emergence of network security in the
mid/late 1990s (again, as a response to a need), and we saw the birth of per-
formance and traffic optimization in the late 1990s/early 2000s, as the
growth of the Internet necessitated optimization and higher-performance
solutions. According to Greg Ness, the result has been “a renaissance of sorts
in the network hardware industry, as enterprises installed successive founda-
tions of specialized gear dedicated to the secure and efficient transport of an
ever increasing population of packets, protocols and services.”
3
Welcome to
the world that has been called Infrastructure1.0 (I-1.0).
The evolution of the basic entity we call I-1.0
is precisely the niche area
that made successful companies such as Cisco, F5 Networks, Juniper, and
Riverbed. I-1.0 established and maintained routes of connectivity between a
globally scaled user base constantly deploying increasingly powerful and
ever more capable network devices. I-1.0’s impact on productivity and com-
merce have been as important to civilization as the development of trans-
oceanic shipping, paved roads, railway systems, electricity, and air travel. I-
1.0 has created and shifted wealth and accelerated technological advance-
ment on a huge number of fronts in countless fields of endeavor. There sim-
ply has been no historical precedent to match the impact that I-1.0 has had
on our world. However, at this point in its evolution, the greatest threat to
the I-1.0 world is the advent of even greater factors of change and complex-
ity as technology continues to evolve. What once was an almost exclusive
domain of firmware and hardware has now evolved to require much more
intelligent and sophisticated software necessary for interfacing with, admin-
istering, configuring, and managing that hardware. By providing such
sophisticated interfaces to firmware/hardware-configured devices, it marked
1. James Urquhart, http://blogs.cisco.com/datacenter/comments/
the_network_the_final_frontier_for_cloud_computing, retrieved 5 Feb 09.
2. http://www.intel.com/technology/mooreslaw/index.htm, retrieved 6 Feb 09.
3. Greg Ness, http://gregness.wordpress.com/2008/10/13/clouds-networks-and-recessions,
retrieved 5 Feb 09.
Chap3.fm Page 58 Friday, May 22, 2009 11:25 AM
The Evolution from the MSP Model to Cloud Computing and Software-as-a-Service
the beginning of the emergence of virtualization. When companies such as
VMware, Microsoft, and Citrix announced plans to move their offerings
into mainstream production data centers such as Exodus Communications,
the turning point for I-1.0 became even more evident. The I-1.0 infrastruc-
ture world was on its way into the cold, dark halls of history.
As the chasm between I-1.0 and the increasingly sophisticated software
packages widened, it became evident that the software could ultimately
drive the emergence of a more dynamic and resilient network. This network
became even more empowered by the addition of application-layer innova-
tions and the integration of static infrastructure with enhanced manage-
ment and connectivity intelligence. The evolving network systems had
become more dynamic and created new problems that software was unpre-
pared to contend with. This gave rise to a new area, virtualization security
(VirtSec, which once again, arose out of necessity), and marked the begin-
ning of an even greater realization that the static infrastructure built over the
previous quarter of a century was not adequate for supporting dynamic sys-
tems or for avoiding the impact that malevolent actions would have on such
dynamic networking paradigms. The recognition that new solutions had to
be developed became apparent when the first virus hit back in the 1970s
(The Creeper virus was first detected on ARPANET, the forerunner of the
Internet, in the early 1970s
4
). No one realized at the time that this single
problem would create an entire industry. As we have discussed, the driving
force for all such technological innovation has been need. For the cloud, the
biggest evolutionary jump began with managed service providers (MSPs)
and their motivation to satisfy and retain customers paying monthly recur-
ring fees.
3.2 The Evolution from the MSP Model to Cloud
Computing and Software-as-a-Service
If you think about how cloud computing really evolved, it won’t take long
to realize that the first iteration of cloud computing can probably be traced
back to the days of frame relay networks. Organizations with frame relay
were essentially singular clouds that were interconnected to other frame
relay-connected organizations using a carrier/provider to transport data
communications between the two entities. Everyone within the frame net-
work sharing a common Private Virtual Connection (PVC) could share
4. http://en.wikipedia.org/wiki/Computer_virus, retrieved 6 Feb 09.
Chap3.fm Page 59 Friday, May 22, 2009 11:25 AM
60 Cloud Computing
their data with everyone else on the same PVC. To go outside their cloud
and connect to another cloud, users had to rely on the I-1.0 infrastructure’s
routers and switches along the way to connect the dots between the clouds.
The endpoint for this route between the clouds and the I-1.0 pathway was a
demarcation point between the cloud and the provider’s customer. Where
the dots ended between the clouds (i.e., the endpoints) was where access
was controlled by I-1.0 devices such as gateways, proxies, and firewalls on
the customer’s premises.
From customers’ perspective, this endpoint was known as the main
point of entry (MPOE) and marked their authorized pathway into their
internal networking infrastructure. By having applications use specific pro-
tocols to transport data (e.g., Simple Mail Transfer Protocol [SMTP] for
sending mail or File Transfer Protocol [FTP] for moving files from one loca-
tion to another), applications behind the MPOE could accept or reject traf-
fic passing over the network and allow email and file transfer to occur with
little to no impedance from the network infrastructure or their administra-
tors. Specialized applications (developed out of necessity to satisfy specific
business needs) often required a client/server implementation using specific
portals created through the firewall to allow their traffic protocols to pro-
ceed unhindered and often required special administrative setup before they
could work properly. While some of this may still hold, that was, for the
most part, how it was done “old school.” Things have changed considerably
since that model was considered state of the art. However state of the art it
was, it was difficult to manage and expensive. Because organizations did not
want to deal with the complexities of managing I-1.0 infrastructure, a cot-
tage industry was born to do just that.
3.2.1 From Single-Purpose Architectures to Multipurpose
Architectures
In the early days of MSPs, the providers would actually go onto customer
sites and perform their services on customer-owned premises. Over time,
these MSPs specialized in implementation of infrastructure and quickly fig-
ured out ways to build out data centers and sell those capabilities off in
small chunks commonly known as monthly recurring services, in addition
to the basic fees charged for ping, power, and pipe (PPP).
Ping
refers to the
ability to have a live Internet connection,
power
is obvious enough, and
pipe
refers to the amount of data throughput that a customer is willing to pay
for. Generally, the PPP part of the charge was built into the provider’s
monthly service fee in addition to their service offerings. Common services
Chap3.fm Page 60 Friday, May 22, 2009 11:25 AM
The Evolution from the MSP Model to Cloud Computing and Software-as-a-Service
provided by MSPs include remote network, desktop and security monitor-
ing, incident response, patch management, and remote data backup, as well
as technical support. An advantage for customers using an MSP is that by
purchasing a defined set of services, MSPs bill a flat or near-fixed monthly
fee, which benefits customers by having a predictable IT cost to budget for
over time. Step forward to today and we find that many MSPs now provide
their services remotely over the Internet rather than having to sell data cen-
ter space and services or perform on-site client visits (which is time-consum-
ing and expensive).
3.2.2 Data Center Virtualization
From the evolutionary growth of the MSP field, coupled with the leaps
made in Internet and networking technology over the past 10 years, we
have come to a point where infrastructure has become almost secondary to
the services offered on such infrastructure. By allowing the infrastructure
to be virtualized and shared across many customers, the providers have
changed their business model to provide remotely managed services at
lower costs, making it attractive to their customers. These X-as-a-Service
models (XaaS) are continually growing and evolving, as we are currently
standing at the forefront of a new era of computing service driven by a
huge surge in demand by both enterprises and individuals. Software-as-a-
Service (SaaS, and other [X]aaS offerings such as IaaS, MaaS, and PaaS)
can be seen as a subset or segment of the cloud computing market that is
growing all the time. One IDC report indicated that cloud computing
spending will increase from $16 billion in 2008 to $42 billion in 2012.
5
Is
there little wonder there is incentive for consumers to pursue cloud com-
puting and SaaS?
Typically, cloud computing has been viewed as a broad array of Internet
Protocol (IP) services (generally using an application called a Web browser
as the main interface) in order to allow users to obtain a specific set of func-
tional capabilities on a “pay for use” basis. Previously, obtaining such ser-
vices required tremendous hardware/software investments and professional
skills that were required in hosting environments such as Exodus Commu-
nications, Cable & Wireless, SAVVIS, and Digital Island. From an enter-
prise customer perspective, the biggest advantages of cloud computing and
5. Roger Smith, “IDC Says IT Cloud Services to Reach $42 Billion by 2012,” http://www.infor-
mationweek.com/blog/main/archives/2008/10/idc_says_it_clo.html, October 2008,
retrieved 6 Feb 2009.
Chap3.fm Page 61 Friday, May 22, 2009 11:25 AM
..................Content has been hidden....................
You can't read the all page of ebook, please click here login for view all page.