Overcoming the Challenges of Today's Cloud Reality
Cloud computing is one of the most hyped approaches to IT today. While it has great potential for cost savings, efficiency and agility, it also introduces some new challenges.
Market adoption models like “Crossing the Chasm” and the “Hype Cycle” indicate cloud computing is gaining momentum and seeing increased adoption. They also suggest that more organizations will encounter challenges—and even failures—in deploying cloud computing environments.
The reality of Cloud today is a combination of high potential and hidden surprises.
Read Overcoming the Challenges of Today's Cloud Reality to understand how cloud computing leaders are building and operating successful cloud environments by making certain that their teams have the necessary resources.
Read more about:
OVERCOMING THE CHALLENGES
OF TODAY’S CLOUD REALITY
May 2011
Rev. B 05/11
SPIRENT
1325 Borregas Avenue
Sunnyvale, CA 94089 USA
Email: sales@spirent.com
Web: www.spirent.com
AMERICAS 1-800-SPIRENT • +1-818-676-2683 • sales@spirent.com
EUROPE AND THE MIDDLE EAST +44 (0) 1293 767979 • emeainfo@spirent.com
ASIA AND THE PACIFIC +86-10-8518-2539 • salesasia@spirent.com
© 2011 Spirent. All Rights Reserved.
All of the company names and/or brand names and/or product names referred to in this
document, in particular, the name “Spirent” and its logo device, are either registered
trademarks or trademarks of Spirent plc and its subsidiaries, pending registration in
accordance with relevant national laws. All other registered trademarks or trademarks are the
property of their respective owners.
The information contained in this document is subject to change without notice and does not
represent a commitment on the part of Spirent. The information in this document is believed
to be accurate and reliable; however, Spirent assumes no responsibility or liability for any
errors or inaccuracies that may appear in the document.
Main Title Style
Subtitle Style
CONTENTS
Executive Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
The State of Cloud Computing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
Moving Ahead Within Today’s Cloud Reality . . . . . . . . . . . . . . . . . . . . . . . . 5
Overcoming the Challenges . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
SPIRENT WHITE PAPER • i
Overcoming the Challenges of Today’s Cloud Reality
1 • SPIRENT WHITE PAPER
EXECUTIVE SUMMARY
Cloud computing is one of the most hyped approaches to IT today. While it
has great potential for cost savings, efficiency and agility, it also introduces
some new challenges. Market adoption models like “Crossing the Chasm” and
the “Hype Cycle” indicate cloud computing is gaining momentum and seeing
increased adoption. They also suggest that more organizations will encounter
challenges—and even failures—in deploying cloud computing environments. The
reality of Cloud today is a combination of high potential and hidden surprises.
Major cloud service interruptions are likely to make headlines, particularly in
this stage of cloud computing adoption. One notable service interruption began
on April 21st, 2011 when Amazon Web Services (AWS) experienced a network
event that resulted in a cascade of issues that impacted customers for several
days. Reports from AWS indicated network congestion caused storage volumes
to be re-mirrored leading to even more network congestion as well as a shortage
of storage resources. Could this have been prevented?
In order to improve the odds of success in deploying an Infrastructure as
a Service (IaaS) environment, leaders of enterprise IT and service provider
organizations must:
• Be familiar with the unique challenges of cloud computing
• Confirm that the right technical and business questions are asked
• Ensure that these questions are answered based on hard data
Proper testing and validation of cloud computing environments is not easy.
Leaders of organizations that are building and operating cloud environments
must make certain their teams have the resources needed for success. These
include:
1. Performance, availability, security, scalability (PASS) testing methodologies
with detailed procedures and test cases
2. Test equipment that supports the latest networking specifications
3. Automated tools designed for cloud testing
Together, these resources can be used to help design, build and validate clouds
that meet the performance, availability, security and scalability requirements of
customers. After all, no one wants to be responsible for a cloud failure.
Overcoming the Challenges of Today’s Cloud Reality
SPIRENT WHITE PAPER • 2
BACKGROUND
Cloud computing is one of
the most popular topics
in the IT world today. Yet,
there is still confusion
on what it really means.
There are at least a couple
of good explanations
for this. First, vendors
and service providers
continue to introduce
competing definitions,
often emphasizing the
unique strengths of their
own offerings. Second,
media organizations tend to over-simplify definitions in hopes of making cloud
computing understandable to the masses. Unfortunately, these approaches
generate more confusion than clarity.
What is needed is an intelligent, vendor neutral explanation of cloud computing
that maintains the substance, breadth and nuances of its meaning. The
Information Technology Laboratory of the National Institute of Standards (NIST)
has done much of that with the following high level definition:
Cloud computing is a model for enabling convenient, on-demand network access
to a shared pool of configurable computing resources (e.g., networks, servers,
storage, applications, and services) that can be rapidly provisioned and released
with minimal management effort or service provider interaction.
The NIST definition goes on to cover some other important elements of cloud
computing by defining three related service models:
• Software as a Service (SaaS) - Applications run on a cloud infrastructure
and are accessible from various client devices, typically via a web browser.
• Platform as a Service (PaaS) - Application development occurs in a
cloud-based development platform and typically enables creation of
web-enabled applications.
• Infrastructure as a Service (IaaS) - Infrastructure components
including servers, storage and networks are available on demand
through network connection.
Many of the challenges in today’s cloud reality are best explained at the
Infrastructure as a Service (IaaS) layer and that is where the remainder of this
white paper will focus. Since some form of Infrastructure as a Service (IaaS) is
often used as a foundation for PaaS and SaaS environments, the issues that will
be discussed still apply to most cloud computing environments.
INTRODUCTION
Leaders across the IT industry are seeking the benefits of cloud computing.
Enterprise IT organizations are building private clouds while service provider
organizations are building public clouds. Enterprise IT organizations are seeking
benefits that include efficiency, cost savings, agility and control. Service
provider organizations are looking to cloud services for new sources of revenue
and competitive advantage, as well as increased efficiency. Cloud computing
seems to be the answer for everything.
Don’t Ignore the Challenges
Of course cloud
computing is not a silver
bullet that solves every
problem faced by IT.
Leaders of enterprise
IT and service provider
organizations are still
responsible for meeting
the needs of all their
stakeholders including
customers, users, business organization, investors. They must deliver on service
level agreements (SLA) and quality of service (QoS) / quality of experience (QoE)
objectives. They must ensure security, performance and reliability. They must also
face the harsh reality that cloud computing introduces new challenges of its own.
Public and private Infrastructure as a Service (IaaS) clouds both create similar
challenges. Intensive utilization of infrastructure resources pushes existing
technology to its limits. Granular resource sharing introduces security threats
to multiple systems and users at the same time rather than one by one. Rapid
changes in consumption levels drive increased competition for resources.
Movement of applications adds complexity to troubleshooting. All these
challenges can dramatically impact the performance, availability, security and
scale of IT services. Those responsible for building and operating Infrastructure
as a Service (IaaS) clouds must understand not only each new challenge, but
also what can be done to overcome them.
“...the harsh reality [is]
that cloud computing
introduces new
challenges of its own.”
THE STATE OF CLOUD COMPUTING
Most would agree that cloud computing has been overhyped for several years.
Yet cloud is real, evidenced by increasing numbers of private cloud deployments
and growing adoption rates for public cloud services. Two well known models for
understanding technology markets provide additional insight to the current state
of cloud computing.
Crossing the Chasm
The Crossing the Chasm model was
introduced and popularized by best-
selling author and consultant Geoffrey
Moore. It begins with the idea that
new technologies are first purchased
by a group called early adopters. They
are willing to endure the problems
associated with fundamentally new
products and services. Then, for
the technology to reach mainstream
adoption, a Chasm must be crossed.
The Chasm is a figurative reference to technologies that fail to meet the needs of
the majority of target buyers. Offerings that fail to cross the Chasm are destined to
serve small markets or simply fail.
Infrastructure as a Service (IaaS) is currently moving from the early adopter phase
to the early majority phase in the Chasm model. In other words, Infrastructure
as a Service (IaaS) is in the process of crossing the Chasm. The good news is that
this indicates Infrastructure as a Service (IaaS) is proving to be valuable and is
gaining traction. The bad news is that there is still a lot of variation in how clouds
are built and few reference models for building clouds properly. This means,
unfortunately, that some clouds will succeed while others will fail.
Hype Cycle
The Hype Cycle was developed by Gartner, a leading IT research and advisory
company. It suggests that new technologies rapidly gain awareness and just
as quickly become over-hyped. When a technology achieves it utmost hype,
it has reached what the model calls the peak of inflated expectations. While
new technologies may have long term potential in line with the hype, they must
first pass through what is called the trough of disillusionment. In this stage,
which begins after the peak of hype is reached, the realities and challenges
associated with the new technology become apparent and opinions become
overly negative.
According to Gartner, the overall notion of cloud computing was hitting the
peak of inflated expectations in 2009. It is now moving through the trough of
disillusionment. As with the chasm model, this stage suggests good news and
bad news. Once again, there is clear value to be found in cloud computing.
However, an increasing number of public failures and other issues will cause
backlash. Rather than just hype, there will also be condemnation and criticism
Proceed with Caution
While both models indicate some concerns, there are also risks associated with
not moving ahead. For instance, relying on old approaches or technology for too
long can lead to competitive gaps or stagnation. The expected rewards from
cloud adoption and the risks of not moving forward indicate continued market
growth. Mainstream adoption is on its way. However, given the current state of
cloud computing, IT leaders should proceed with caution
MOVING AHEAD WITHIN TODAY’S CLOUD REALITY
As mentioned, enterprise IT organizations and service providers have some
slightly different reasons for moving to cloud. Similarly, the benefits they seek—
for themselves and for their users—are also slightly different. Whether you are
building and operating a private cloud or a public cloud, it is important to clarify
the expected benefits for users and customers.
Business Drivers and Benefits
Service Providers
For service providers, the addition of cloud computing services may be done for
several different reasons. Managed service providers (MSP) use Infrastructure
as a Service (IaaS) clouds to increase efficiency, lower costs and increase
profits. MSPs often provide fixed price contracts to their customers, so
increases in efficiency during the life of a given contract flow straight to their
bottom line. Hosting providers are adding Infrastructure as a Service (IaaS)
in response to new competitors that are focused exclusively cloud services.
Communications service providers (CSP) are leveraging existing investments in
network infrastructure to gain a new source of revenue.
Regardless of why service providers build clouds, they must deliver value to their
customers. By supporting new business models such as pay-per-use, service
providers let their customers shift capital investments to expenses. By pooling large
amounts of resources, service providers give customers access to what appears to
be an unlimited amount of IT capacity. Service providers also take responsibility for
space, power, cooling and certain IT management tasks, performing those activities
more efficiently than individual customers could on their own.
Enterprise IT
Many enterprise Infrastructure as a Service (IaaS) cloud deployments are
initially designed to save money. Physical servers have been under-utilized,
typically just seeing 10-15% average CPU consumption rates. This has driven IT
organizations to consolidate servers using virtualization, bringing average CPU
consumption closer to 40% or even 50% utilization. Cloud computing is now
seen as the next step in efficiency and can produce utilization levels as high as
75% or more.
Users of enterprise IT clouds enjoy many of the benefits of public clouds and
cite some additional advantages. Accurate or not, some organizations simply
feel security is greater when infrastructure, applications and data are located
on premise. When digging deeper into this perspective, the underlying concern
often turns out to be a desire for control rather than specific risks. In any case,
private clouds are a viable alternative. They also serve as the foundation for
hybrid clouds where certain applications and data always remain on premise
and other applications may access public resources in times of peak demand.
The Benefit of Agility
Agility is sometimes a “surprise” benefit. IT leaders may initially focus on the
financial benefits of cloud computing. Then, once users are taking advantage
of the cloud services, agility is often listed as the top outcome. Both public and
private clouds let their users become far more agile. For example, users get self-
service access to IT resources the moment they are needed. This eliminates a
long series of delays involving purchase orders, approvals, shipment and setup.
Without these delays, users can more quickly adapt to changes in demand or in
the competitive environment.
Infrastructure as a Service (IaaS) is not just Server Virtualization
Cloud computing provides some obvious benefits. Still, challenges remain.
One source of problems stems from equating Infrastructure as a Service (IaaS)
with server virtualization. Any organization can deploy virtualized servers.
Enterprises have done plenty of this through server consolidation. Similarly,
hosting providers have offered virtual private servers (VPS) for years. However,
it is important to realize that Infrastructure as a Service (IaaS) is not simply
server virtualization.
Infrastructure as a Service (IaaS) adds characteristics such as on-demand self-
service, allowing users to provision resources by leveraging automation rather
than involving administrators. Further, large amounts of resources are pooled
together and shared in Infrastructure as a Service (IaaS) environments. Resource
pooling enables rapid elasticity or the ability to quickly gain access to additional
resources and just as quickly return them for use by others. Infrastructure as
a Service (IaaS) clouds may also support application awareness and dynamic
resource optimization. This is where building and operating cloud environments
becomes more difficult.
High-level Challenges
Increased resource sharing creates a number of challenges. When done properly,
it means higher utilization and greater efficiency. When done poorly it leads to
resource contention and failed SLAs. Resource sharing also requires diligent
enforcement of security. Applications using the same server or database must
not have access to data from other applications. Cloud providers must also
ensure that noisy neighbors—applications attempting to consume more than
their fair share of resources—do not impact other applications.
Cloud providers have plenty of other challenges to consider. Infrastructure as a
Service (IaaS) environments tend to have an API which provides programmatic
control over resources. This can lead to extremely fast changes in resource
consumption as well as rapid modifications to infrastructure configurations.
Cloud providers must have intelligence and automation to move virtual
machines (VM) from one server to another in response to changes in demand
and resource utilization. All this must be handled transparently to the
applications that are running on highly utilized servers with densely packed
VMs.
Cloud environments tend to have a greater reliance on virtualized resources
with server virtualization as the most obvious. Another example is storage
virtualization through SANs which enables VMs to run on any server. Network
virtualization including virtual switches, load balancers and firewalls is also
predominant. Even applications themselves are becoming more virtualized.
For example, as the web tier or application tier encounters more demand,
elastic applications can self-provision additional resources to maintain SLA
commitments. Of course virtual components have different performance and
scalability attributes which adds a number of design challenges.
Technology Decisions
At some point, the right technology decisions must be made in order to
overcome the challenges and deliver the value promised by cloud computing.
Not surprisingly, architecture, design and implementation decisions have a
huge impact on how cloud environments behave. This includes performance,
availability, scale and security, also known as PASS.
Making the right technical decisions requires many complex questions to be
answered accurately:
• How many 10Gbps connections can my servers potentially saturate?
• Is 10Gbps or 40Gbps Ethernet the best choice?
• What about Fibre Channel over Ethernet (FCoE) for a converged LAN / SAN?
• Is a non-blocking SAN required or can it be over-provisioned?
• Does one vendor or another have a better core network fabric solution?
• Should low-end 8-core servers or high-end 20-core servers be used?
• Which configuration supports more VMs or generates the most revenue?
• How well does a particular hypervisor / server combination work?
• What does a shared virtual switch mean to server performance?
• How does the security of a software appliance compare to a hardware device?
• Are security profiles properly moving along with the VMs?
Depending on the requirements of a given cloud environment, there may be
many more considerations. These technical questions raise an even more
important question for IT leaders. “Does your team have the appropriate tools
for answering all these questions?”
More Questions for IT Leaders
While architects and other senior technical staff must ultimately answer the
technical questions, IT managers, including the CIO, remain responsible for the
overall outcome of cloud implementations. IT managers are accountable for
ensuring good answers to all of these questions:
• Has the team come up with the right architecture?
• Are we using the right components?
• Will we meet performance and SLA requirements once it is built?
• What will happen during periods of peak demand?
• How will we determine the underlying cause when something goes wrong?
IT managers do not want to become victims of the Chasm or the trough of
disillusionment. They want to realize the benefits of cloud computing and reap
the accompanying rewards. To do this, they must ensure that the answers to all
of these important questions are based on hard data.
OVERCOMING THE CHALLENGES
The hard data needed to ensure success comes from PASS testing which
includes assessments of performance, availability, security and scalability.
PASS testing is done on infrastructure components such as servers, network
elements, storage devices and virtual appliances. It is also performed at the
system level to ensure the assembled components still perform correctly when
used together by applications.
PASS Testing for Cloud Computing Environments
As discussed, cloud computing environments are
different than traditional datacenter environments
and create an array of new challenges. Some of the
challenges show up in PASS testing
Performance
Measuring the performance of a cloud computing
environment is not sufficiently done with one or
even a dozen tests. Many hundreds of individual
measurements—including throughput and latency
of virtual switches, virtual appliances, converged
network adapters (CNA), top of rack switches,
datacenter cross-connects and various applications—
are required to fully characterize cloud performance.
Then, a good portion of these tests must be repeated
under a variety of different conditions. What happens,
for instance, when server and storage traffic are
simultaneously reaching peaks? Will virtual switches
within hypervisors keep up with high network traffic when applications are
competing for the same CPU?
Availability
Availability in Infrastructure as a Service (IaaS) clouds is enhanced by
capabilities such as live migration, dynamic resource scheduling and automatic
restarts. Yet these same features have the potential to create problems of their
own. Since live migrations happen when load on the infrastructure is high,
they place additional demand on resources that are already scarce. The net
result may actually be further performance or availability degradation. PASS
testing must be used to measure VM downtime during live migrations. Testing
should include moving VMs, along with their running applications, from heavily
loaded servers. It should also include measuring failover time for dozens of
simultaneous high availability restarts due to a physical server failure.
http://www.spirent.com/
White-Papers/Broadband/
PAB/EANTC_CloudComputing_
Whitepaper.aspx
Security
Multi-tier applications—consisting of web, application and database tiers—
produce network traffic multiple times during each transaction. Since security is
required at every layer, including the hypervisor, virtual and physical appliances
may impact application performance. Attempts at tuning performance may
result in accidental vulnerabilities. With cloud computing, security must address
targeted attacks, viruses and other risks while maintaining the performance,
availability and scalability for all cloud tenants. PASS testing helps uncover
vulnerabilities and can also be used to determine the right mix of physical and
virtual security appliances. This is an important consideration since every CPU
cycle spent checking security cannot be sold to customers.
Scalability
Cloud environments tend to be quite large, from hundreds of servers to tens of
thousands. Consider a cloud infrastructure with a thousand servers. It will have
many thousands of cores supporting tens of thousands of VMs. This type of
density leads to a huge increase in network traffic. Scalability testing must not
only validate that tens of thousands of VMs can operate simultaneously, it must
confirm they can do so with peak traffic rates across all layers of the network.
Tools and Procedures
Homegrown tools and scripts are insufficient for testing and validating
complex systems such as cloud computing environments. In fact, testing in
general typically requires automated
test tools that are just as complex and
powerful as the system under test.
The selection of commercial testing
products must be done carefully. Cloud
networks take advantage of the newest
equipment, built with the latest network
specifications. Be sure to select testing
products that are continually updated to
support leading edge cloud networks.
Test cases and procedures must also
be properly developed in order to
accomplish the desired results. The
Spirent Journal of PASS Test Methodologies by Spirent Communications is an
element of the Spirent test ecosystem that defines and documents the most
critical PASS test cases. It includes the Spirent Catalogue of Test Methodologies
which are intended to help development engineers and product verification
engineers rapidly develop and test complex scenarios. Sections like “Testing
Cloud Application and Security Services” help clarify what should be tested and
provide step-by-step procedures for doing so.
Using PASS test methodologies along with automated tools and test cases
greatly improves the odds of success for any cloud environment. Hard data
from PASS testing should be gathered and used during the architecture and
design phase, deployment phase and during the ongoing operation of the cloud
environment. A faulty assumption on virtual switch performance should not
be the cause of a bad architecture. When problems do arise in an operational
cloud, a single mis-configured box must be quickly found among thousands and
fixed before widespread SLA violations occur.
CONCLUSION
Cloud computing has great potential for cost
savings, efficiency, agility and more. These
compelling benefits are driving higher adoption
rates for public cloud services and increased
deployments of private clouds.
Models like Crossing the Chasm and the Hype
Cycle indicate cloud computing is in the process
of moving toward mainstream adoption. They
also suggest that more organizations will
encounter challenges—and even failures—in
deploying cloud computing environments. The
reality of cloud today is a combination of high potential and hidden surprises.
In order to improve the odds of success in deploying an IaaS environment,
leaders of enterprise IT and service provider organizations must be familiar with
the unique challenges of cloud computing. They must confirm that the right
technical and business questions are asked. They must also make certain that
these questions are properly answered, based on hard data derived from testing
and validation.
Proper testing and validation of cloud computing environments is not easy.
Leaders of organizations that are building and operating cloud environments
must ensure their teams have the resources needed for success. The three key
points to remember are:
1. PASS testing methodologies with detailed procedures and test cases
2. Test equipment that supports the latest networking specifications
3. Automated tools designed for cloud testing
www.spirent.com/testedwithspirent
