UNIT-III

Cloud Computing Architecture and

Management
 Cloud architecture
• The cloud architecture can be divided into
four layers based on the access of the cloud by
the user.
– Layer 1 (User/Client Layer)
– Layer 2 (Network Layer)
– Layer 3 (Cloud Management Layer)
– Layer 4 (Hardware Resource Layer)
 Layer 1 (User/Client Layer)
• This layer is the lowest layer in the cloud architecture. All the
users or client belong to this layer.
• This is the place where the client/user initiates the connection
to the cloud. The client can be any device such as a thin client,
thick client, or mobile or any handheld device that would
support basic functionalities to access a web application.
• The thin client here refers to a device that is completely
dependent on some other system for its complete functionality.
• thick clients are general computers that have adequate
processing capability.
• Usually, a cloud application can be accessed in the same way as
a web application. But internally, the properties of cloud
applications are significantly different. Thus, this layer consists
of client devices.
 Layer 2 (Network Layer)

• This layer allows the users to connect to the cloud.

• The whole cloud infrastructure is dependent on this connection where
the services are offered to the customers.
• This is primarily the Internet in the case of a public cloud. The public
cloud usually exists in a specific location and the user would not know
the location as it is abstract. And, the public cloud can be accessed all
over the world.
• In the case of a private cloud, the connectivity may be provided by a
local area network (LAN). Even in this case, the cloud completely
depends on the network that is used.
• Usually, when accessing the public or private cloud, the users require
minimum bandwidth, which is sometimes defined by the cloud
providers.
• This layer does not come under the purview of service-level
agreements (SLAs), that is, SLAs do not take into account the Internet
 Layer 3 (Cloud Management Layer)
• This layer consists of software that are used in managing the cloud. The
software can be a cloud operating system (OS).
• Software that acts as an interface between the data center (actual resources)
and the user, or a management software that allows managing resources.
• These software usually allow resource management (scheduling, provisioning,
etc.), optimization (server consolidation, storage workload consolidation), and
internal cloud governance.
• This layer comes under the purview of SLAs, that is, the operations taking place
in this layer would affect the SLAs that are being decided upon between the
users and the service providers. Any delay in processing or any discrepancy in
service provisioning may lead to an SLA violation. As per rules, any SLA
violation would result in a penalty to be given by the service provider.
• These SLAs are for both private and public clouds. Popular service providers are
Amazon Web Services (AWS) and Microsoft Azure for public cloud. Similarly,
OpenStack and Eucalyptus allow private cloud creation, deployment, and
management.
 Layer 4 (Hardware Resource Layer)

• Layer 4 consists of provisions for actual hardware resources. Usually, in

the case of a public cloud, a data center is used in the back end.
Similarly, in a private cloud, it can be a data center, which is a huge
collection of hardware resources interconnected to each other that is
present in a specific location or a high configuration system.
• This layer comes under the purview of SLAs. This is the most important
layer that governs the SLAs. This layer affects the SLAs most in the case
of data centers.
• Whenever a user accesses the cloud, it should be available to the users
as quickly as possible and should be within the time that is defined by
the SLAs.
• As mentioned, if there is any discrepancy in provisioning the resources or
application, the service provider has to pay the penalty. Hence, the data
center consists of a high-speed network connection and a highly efficient
algorithm to transfer the data from the data center to the manager.
 Anatomy of the Cloud
• Cloud anatomy can be simply defined as the structure of the cloud.
• Architecture is a hierarchical structural view that defines the
technology as well as the technology over which it is dependent
or/and the technology that are dependent on it. Thus, anatomy can
be considered as a part of architecture.
• Standard anatomy that is the base for the cloud. It depends on the
person to choose the depth of description of the cloud.
• There are basically five components of the cloud:
– 1. Application: The upper layer is the application layer. In this layer, any
applications are executed.
– 2. Platform: This component consists of platforms that are responsible for
the execution of the application. This platform is between the
infrastructure and the application.
– 3. Infrastructure: The infrastructure consists of resources over which the
other components work. This provides computational capability to the
user.
– 4. Virtualization: Virtualization is the process of making logical
components of resources over the existing physical resources. The logical
components are isolated and independent, which form the infrastructure.
– 5. Physical hardware: The physical hardware is provided by server and
storage units.
Network Connectivity in Cloud Computing
• Cloud computing is a technique of resource sharing where
servers, storage, and other computing infrastructure in multiple
locations are connected by networks. In the cloud, when an
application is submitted for its execution, needy and suitable
resources are allocated from this collection of resources; as
these resources are connected via the Internet, the users get
their required results.
• For many cloud computing applications, network performance
will be the key issue to cloud computing performance.
– Public Cloud Access Networking
– Private Cloud Access Networking
– Intracloud Networking for Public Cloud Services
– Private Intracloud Networking
– New Facets in Private Networks
 Public Cloud Access Networking

• The connectivity is through the Internet, though some cloud providers

may be able to support virtual private networks (VPNs) for customers.
• Accessing public cloud services will always create issues related to
security, which in turn is related to performance.
• One of the possible approaches toward the support of security is to
promote connectivity through encrypted tunnels, so that the
information may be sent via secure pipes on the Internet.
• This procedure will be an overhead in the connectivity, and using it
will certainly increase delay and may impact performance.
• If we want to reduce the delay without compromising security, then
we have to select a suitable routing method such as the one reducing
the delay by minimizing transit hops in the end-to-end connectivity
between the cloud provider and cloud consumer
 Private Cloud Access Networking
• In the private cloud deployment model, since the
cloud is part of an organizational network, the
technology and approaches are local to the in-
house network structure.
• This may include an Internet VPN or VPN service
from a network operator.
• If the application access was properly done with an
organizational network—connectivity in a precloud
configuration—transition to private cloud
computing will not affect the access performance
 Intracloud Networking for Public Cloud Services

• Here, the resources of the cloud provider and thus the

cloud service to the customer are based on the resources
that are geographically apart from each other but still
connected via the Internet.
• Public cloud computing networks are internal to the
service provider and thus not visible to the user/customer;
• The security aspects of connectivity and the access
mechanisms of the resources and the QoS in the
connected resources worldwide.
• Most of the performance issues and violations from these
are addressed in the SLAs commercially.
 Private Intracloud Networking
• The most complicated issue for networking and connectivity in cloud
computing is private intracloud networking. It depends on how much
intracloud connectivity is associated with the applications being executed in
this environment.
• Private intracloud networking is usually supported over connectivity between
the major data centre sites owned by the company.
• At a minimum, all cloud computing implementations will rely on intracloud
networking to link users with the resource to which their application was
assigned.
• Once the resource linkage is made, the extent to which intracloud networking
is used depends on whether the application is componentized based on
service-oriented architecture (SOA) or not, among multiple systems.
• There are reasons to consider the networks and connectivity in cloud
computing with newer approaches as globalization and changing network
requirements, especially those related to increased Internet usage, are
demanding more flexibility in the network architectures of today’s enterprises.
 New Facets in Private Networks
• Conventional private networks have been architected for on-premise
applications and maximum Internet security.
• Typically, applications such as e-mail, file sharing, and enterprise
resource planning (ERP) systems are delivered to on-premise-based
servers at each corporate data center.
• Increasingly today, software vendors are offering Software as a
Service (SaaS) as an alternative for their software support to the
corporate offices, which brings more challenges in the access and
usage mechanisms of software from data center servers and in the
connectivity of network architectures.
• The traditional network architecture for these global enterprises was
not designed to optimize performance for cloud applications, now
that many applications including mission-critical applications are
transitioning (moving) from on-premise based to cloud based,the
business cannot function if it cannot access applications such as ERP
 Path for Internet Traffic

• The traditional Internet traffic through a limited set of Internet gateways

poses performance and availability issues for end users who are using
cloud-based applications.
• It can be improved if a more widely distributed Internet gateway
infrastructure and connectivity are being supported for accessing
applications, as they will provide lower-latency access to their cloud
applications.
• As the volume of traffic to cloud applications grows, the percentage of
the legacy network’s capacity in terms of traffic to regional gateways
increases.
• Applications such as video conferencing would need more bandwidth
while applications such as ERP(Enterprice resource planning) will
consume less bandwidth, and hence, one has to plan a correct
connectivity and path between providers and consumers.
 Applications
• A cloud application usually can be accessed as
a web application
•

Computer application evolution.

• Multitenancy
• Elasticity
• Heterogeneous cloud platform
• Quantitative measurement
• On-demand service
 Managing the Cloud
• Cloud management is aimed at efficiently
managing the cloud so as to maintain the QoS.
It is one of the prime jobs to be considered.
The whole cloud is dependent on the way it is
managed. Cloud management can be divided
into two parts:
– 1. Managing the infrastructure of the cloud
– 2. Managing the cloud application
 Migrating Application to Cloud
• Cloud migration encompasses moving one or more
enterprise applications and their IT environments from
the traditional hosting type to the cloud environment,
either public, private, or hybrid.
• Cloud migration presents an opportunity to significantly
reduce costs incurred on applications.
• Phases of cloud migration are
– Evaluation
– Migration strategy
– Prototyping
– Provisioning
– Testing
 Phases of Cloud Migration
1.Evaluation:
– Evaluation is carried out for all the components like current infrastructure
and application architecture, environment in terms of compute, storage,
monitoring, and management, SLAs, operational processes, financial
considerations, risk, security, compliance, and licensing needs are identified
to build a business case for moving to the cloud.
2.Migration strategy:
Based on the evaluation, a migration strategy is drawn—a hotplug strategy
is used where the applications and their data and interface dependencies
are isolated and these applications can be operationalized all at once.
A fusion strategy is used where the applications can be partially migrated;
but for a portion of it, there are dependencies based on existing licenses,
specialized server requirements like mainframes, or extensive
interconnections with other applications
3.Prototyping:
Migration activity is preceded by a prototyping activity to validate and
ensure that a small portion of the applications are tested on the cloud
environment with test data setup.
4. Provisioning:
Cloud servers are provisioned for all the identified environments, necessary
platform softwares and applications are deployed, configurations are tuned
to match the new environment sizing, and databases and files are replicated.
All internal and external integration points are properly configured. Web
services, batch jobs, and operation and management software are set up in
the new environments.
5. Testing:
Postmigration tests are conducted to ensure that migration has been
successful.
Performance and load testing, failure and recovery testing, and scale-out
testing are conducted against the expected traffic load and resource
utilization levels.
 Approaches for Cloud Migration
• The following are the four broad approaches
for cloud migration that have been adopted
effectively by vendors:
1. Migrate existing applications
2. Start from scratch
3. Separate company
4. Buy an existing cloud vendor
1. Migrate existing applications:
• Rebuild or rearchitect some or all the applications, taking advantage of
some of the virtualization technologies around to accelerate the work.
2. Start from scratch:
Rather than cannibalize sales, confuse customers with choice, and tie up
engineers trying to rebuild existing application, it may be easier to start
again
3. Separate company: One may want to create a whole new company with
separate brand, management, R&D, and sales. The investment and
internet protocol (IP) may come from the existing company. The separate
company may even be a subsidiary of the existing company.
4. Buy an existing cloud vendor:
For a large established vendor, buying a cloud-based competitor achieves
two things.
Firstly, it removes a competitor
Secondly, it enables the vendor to hit the ground running in
the cloud space.
The risk of course is that the innovation, drive, and operational approach
of the cloud-based company are destroyed as it is merged into the larger
acquirer