MODULE-IV

NETWORKING FOR CLOUD COMPUTING

Introduction
The Internet has grown into a huge source of information, with data centers at its center.
These centers provide services (like Google, Facebook) and resources (like Amazon EC2,
Microsoft Azure). With the rise of cloud computing, data centers are now essential for
businesses. They help manage data and IT infrastructure but are complex and expensive to
maintain.

Data centers are classified into four tiers based on how reliable they are. Tier IV is the most
reliable, designed for critical systems, while Tier I is for smaller businesses. Because of
high costs, small companies often rely on cloud service providers like Google and Amazon.

As cloud services grow, data centers face challenges in cutting costs, saving energy, and
improving performance. This chapter looks at these challenges, especially in data center
networks.
Overview of Data Center Environment
Initially, organizations relied on server rooms to store servers and networking equipment
for LANs. Some had a main server room and a backup in another building for fault
tolerance.

With the rise of client-server computing and the Internet, the concept of data centers
emerged. These centers became vital during the dot-com era, offering high reliability to
support the rapid growth of Internet companies. Over time, data centers became the
backbone of IT, essential for running businesses or serving as the business itself.

Data centers are known for their "mission-critical" role, ensuring maximum uptime. They
are built for reliability at any cost. However, in the past 5–10 years, advancements have
transformed and accelerated innovation in data center technology.

Architecture of Classical Data Centers

Classical data centers store the computational power, storage, and applications essential
for enterprise operations, serving as the core of IT architecture. Proper planning of their
infrastructure is vital to ensure performance, scalability, and resilience. The most common
design is the multitier model, which organizes servers into web, application, and database
layers to support enterprise solutions like ERP and CRM systems. These layers, often used
in environments like Microsoft .NET or Java 2 Enterprise Edition, rely on network-provided
security and application optimization.
In the multitier model, HTTP-based applications are divided into three tiers:

1. Web server
2. Application
3. Database

Another design, the server cluster model, originated in scientific and academic fields and
is now used in industries like finance, manufacturing, and entertainment. It supports high-
performance, parallel, and high-throughput computing, often customized to meet specific
business needs.

These tiers can run on the same machine using interprocess communication or across
separate machines via network connections. Separating tiers enhances resiliency,
allowing servers to be replaced without disrupting operations, and improves security by
isolating compromised web servers from application or database tiers. Typically, web and
application servers may share physical hardware, while databases remain separate.

CEDCs

New technological and economic pressures are forcing organizations to maximize their IT
infrastructure, as the current environment is strained, making it harder to maintain
efficiency and service quality, regardless of the number of users or applications. The
solution lies in cloud computing, which reduces costs and increases flexibility for
applications and IT infrastructure. The Cloud-Enabled Data Center (CEDC) takes
virtualization to the next level, transforming traditional virtualized environments into more
agile systems. It incorporates intelligent, integrated Infrastructure-as-a-Service (IaaS) and
Platform-as-a-Service (PaaS) to manage dynamic workloads, providing resources based on
business policies, while offering automation and coordination of resources across various
data centers. This progression from virtualization management to CEDC addresses
common business goals, improving overall efficiency.

Physical Organization

Data centers can occupy an entire building, a few floors, or even just a single room. They
typically consist of many servers mounted in rack cabinets arranged in rows, creating
aisles for easy access to the front and rear of each cabinet.
Some equipment, like storage devices, can be as large as the racks and are often placed
beside them. In large data centers, which can store thousands of servers, shipping
containers holding 1,000 or more servers may be used. When a failure occurs or upgrades
are needed, entire containers are replaced instead of individual servers.

Storage and Networking Infrastructure

Data centers typically require four types of network access, each using different physical
networks:

1. Client-server network: Provides external connectivity to the data center, using

traditional wired Ethernet or Wireless LAN technologies.
2. Server-server network: Enables high-speed communication between servers,
using Ethernet, InfiniBand, or other technologies.
3. Server-storage network: Connects servers to storage devices with high-speed
links, typically using Fiber Channel, though Ethernet or InfiniBand can also be used.
4. Management network: Manages the data center, usually using Ethernet with
different cabling from the main networks.

Multiple applications run inside a data center, typically with each application hosted on its
own set of (potentially virtual) server machines. Each application is associated with one or
more publicly visible and routable IP addresses, which clients on the Internet use to send
requests and receive replies. Inside the data center, requests are spread among a pool of
frontend servers by a specialized load balancer.
A two-tier network topology is commonly used in data center networks today. Access
switches for server connectivity are collapsed into high-density aggregation switches,
which handle the switching and routing for access switching interconnections and various
LAN servers. It offers several benefits:

• Design simplicity (fewer switches and fewer managed nodes)

• Reduced network latency (fewer switch hops)
• Typically, a reduced network design oversubscription ratio
• Lower aggregate power consumption

However, the two-tier design has limited scalability. When the ports on an aggregation
switch pair are fully utilized, adding another pair of aggregation switches introduces
complexity. The connection between aggregation switch pairs must be fully meshed with
high bandwidth to avoid bottlenecks. Since the aggregation switch pair also runs routing
protocols, more switch pairs mean more routing protocol peering and more routing
interfaces, adding complexity to the full mesh design.

Cooling Infrastructure
Since data centers typically span large areas and house thousands of servers, a
sophisticated cooling system is required, often including building-level air-conditioning
units, fans, and air recirculation systems. The server racks are placed on raised platforms
with alternating back-facing and front-facing aisles. Cold air is forced into the front-facing
aisles, and server fans draw the cold air through the servers, pushing the hot air to the
back. This hot air rises and is directed toward the chiller plant for cooling and recirculation.
While this setup is cost-effective, it can create hot spots due to uneven cooling or the
mixing of hot and cold air.

Cooling in a data center

Recent innovations in power, cooling technologies, and facility management have

improved efficiency in every aspect of data center and building design, incorporating green
technologies and environmental awareness. The data center model has evolved, with
options now including ownership, colocation, wholesale facilities, and public/private
cloud or hybrid strategies.The definition of a data center has changed, with companies like
Google, Microsoft, and Yahoo operating hyperscale facilities with advanced engineering.
Consolidation projects have transformed traditional data centers into a few large-scale
facilities, while the need for more power and cooling has led to the development of more
advanced infrastructure. Site selection for data centers has been influenced by the spread
of fiber networks, the need to avoid natural disasters, and the pursuit of extreme power
and cooling efficiencies. In the U.S., data center hubs have emerged in regions like Silicon
Valley, Quincy (Washington), Chicago, Dallas/Fort Worth, North Carolina, and the New
York/New Jersey area.

Nature of Traffic in Data Centers

The traffic in data center networks (DCNs) differs significantly from that on the internet,
with a much lower round-trip time (RTT) in DCNs—often as low as 250 μs without queuing.
This is because DCNs are private networks designed for high bandwidth and low latency.
The traffic in DCNs is classified into three main types:

1. Mice traffic: Small query traffic, such as Google searches or Facebook updates. It
makes up most of the traffic in a DCN but involves low data transmission volumes.
2. Cat traffic: Control and coordination messages, such as small- to medium-sized
file downloads.
3. Elephant traffic: Large data transfers, like antivirus updates or movie downloads.

These types of traffic—bursty query traffic, delay-sensitive cat traffic, and throughput-
sensitive elephant traffic—often coexist in DCNs, each with its own application and
performance requirements.

Networking Issues in Data Centers

Cloud networking is the backbone of cloud services, driving the transformation of IT
service delivery. This section highlights key challenges in cloud networking.

Availability

A major challenge for cloud providers is ensuring maximum uptime. Even short downtime
can harm the organization’s reputation, breach SLAs, and impact revenue. Downtime not
only disrupts user services but also erodes trust, making reliability critical for business
continuity. High availability is typically achieved through strategies like data replication,
load balancing, and regular backups to minimize service interruptions.

Poor Network Performance

Data Center Networks (DCNs) require high burst tolerance, low latency, and high
throughput to handle diverse traffic types: mice, cat, and elephant traffic, each with unique
performance needs. However, traditional TCP/IP, designed for internet scenarios,
struggles to deliver optimal performance in DCNs.

Security

Ensuring data security during transmission and at rest remains a key concern for cloud
providers. Accidental data loss due to power outages or backup failures can result in
significant damage. Additionally, providers must safeguard the physical security of data
center buildings and networking infrastructure to prevent attacks from malicious insiders.

Transport Layer Issues in DCNs

TCP, a transport protocol, handles most traffic in Data Center Networks (DCNs) and has
been key to the Internet's success due to its effective congestion control mechanisms.
While TCP has evolved to meet growing network demands, it struggles to meet DCN
requirements—high burst tolerance, low latency, and high throughput—simultaneously.
Challenges like TCP incast, TCP outcast, queue buildup, buffer pressure, and pseudo-
congestion hinder its performance, as discussed in the following sections.

TCP Impairments in DCNs

Despite decades of evolution, TCP faces challenges due to the diversity of modern
networks and application demands. These issues expose limitations in TCP’s fundamental
design, particularly in its congestion control mechanisms. This section highlights the key
challenges encountered by state-of-the-art TCP in DCNs.

TCP Incast

TCP incast refers to the issue where TCP fails to fully utilize link capacity in many-to-one
communication patterns, such as those used in web search, MapReduce, and social
networks. This problem is common in large-scale applications and data centers, including
distributed storage and data-intensive computing systems.

In many-to-one patterns, an aggregator sends data requests to multiple worker nodes,

which then send large amounts of data back. When all the worker nodes transmit data
simultaneously, the bottleneck link may overflow, causing packet losses. This is due to
synchronized packet transmission, where workers wait for all nodes to finish sending data
before continuing.
 TCP Incast

As the number of concurrent worker nodes increases, packet losses rise, reducing
throughput. Lost packets are only retransmitted after a delay, affecting time-sensitive
applications like mice traffic. Fast retransmit mechanisms often fail for mice traffic due to
the small data volume and lack of sufficient duplicate acknowledgments.

To address TCP incast, various solutions have been proposed, from modifying TCP itself to
exploring fixes at the application, transport, or link layer. Ren et al. provide an in-depth
analysis of these solutions.
TCP Outcast

TCP outcast occurs when a large set of flows and a small set of flows compete for the
same bottleneck output port in data center switches, causing the small flows to lose their
throughput share. This happens due to drop-tail queues in switches, which lead to
consecutive packet drops and frequent TCP timeouts, resulting in high latency and poor
response times. This issue, known as port blackout, significantly affects the performance
of small flows.

Example Scenario of port blackout

Additionally, TCP’s behavior causes RTT bias, where flows with low RTT get a larger share
of bandwidth than those with high RTT. However, in TCP outcast scenarios, the low RTT
flows are outcasted by the high RTT flows, causing inverse RTT bias.

The two main causes of TCP outcast are the use of drop-tail queues and many-to-one
communication patterns, which are common in data centers.

Mitigating TCP Outcast: To mitigate TCP outcast, alternatives to drop-tail queues can be
used, such as Random Early Detection (RED) or Stochastic Fair Queueing (SFQ). Another
approach is to design efficient TCP congestion control mechanisms at end hosts to reduce
buffer occupancy at switches.
Queue Buildup

In Data Center Networks (DCNs), mice traffic, cat traffic, and elephant traffic coexist.
Elephant traffic, being long-lasting and greedy, often leads to extreme congestion,
overflowing bottleneck buffers. When both mice and elephant traffic share the same route,
the performance of mice traffic is severely impacted.

There are two main ways elephant traffic degrades mice traffic performance:

1. Elephant traffic fills up most of the buffer, increasing the likelihood of packet drops
for mice traffic, similar to the TCP incast issue. Frequent packet losses lead to
timeouts and reduced performance.
2. Even when no packets are lost, mice traffic experiences increased queuing delays
as they wait behind elephant traffic, leading to queue buildup.

Mitigating Queue Buildup: To address queue buildup, the key is to minimize queue
occupancy in DCN switches. Most existing TCP variants use a reactive approach, adjusting
the sending rate only after packet loss occurs. A proactive approach is needed to reduce
queue occupancy and prevent queue buildup.

Buffer Pressure

Buffer pressure occurs due to the greedy and long-lasting nature of elephant traffic. When
both mice and elephant traffic share the same route, elephant traffic occupies most of the
buffer space, leaving little room for the bursty mice traffic. This results in a high number of
packet losses for mice traffic, leading to poor performance. Additionally, since most traffic
in DCNs is bursty, mice traffic packets are frequently dropped because elephant traffic
keeps the buffer space full.

Mitigating Buffer Pressure: To reduce buffer pressure, minimizing buffer occupancy in

switches is crucial, just like with the queue buildup problem.

Pseudocongestion Effect

Virtualization, a key technology in cloud computing, allows data centers to run virtual
machines (VMs) to provide on-demand services. While virtualization offers benefits like
efficient server utilization and lower maintenance costs, it can negatively impact
traditional protocols such as TCP and UDP. A study on Amazon EC2 showed that
virtualization reduces TCP and UDP performance, causing instability in throughput and
increased end-to-end delays.

When multiple VMs run on the same physical machine, hypervisor scheduling latency
increases, leading to longer waiting times for each VM to access the processor. This
latency, ranging from microseconds to hundreds of milliseconds, causes unpredictable
network delays, affecting throughput and increasing delays. If the latency is high enough, it
can trigger a Retransmission Timeout (RTO), causing the VM sender to mistakenly assume
network congestion and lower its sending rate. This phenomenon is known as
pseudocongestion.

Mitigating Pseudocongestion Effect: To address pseudocongestion, two approaches are

suggested:

1. Designing efficient hypervisor schedulers to reduce scheduling latency.

2. Modifying TCP to intelligently detect pseudocongestion and react accordingly.

Cloud Service Providers

Introduction

Cloud computing is a widely discussed technology that provides resources to users as

services like software, infrastructure, platforms, and security. These services are delivered
via the Internet from cloud providers rather than on-premise servers.
Cloud services offer easy, scalable access to applications, resources, and services, fully
managed by providers. They dynamically scale to meet user needs, eliminating the need
for companies to maintain their own IT resources or staff. Examples include online storage,
web-based email, office tools, database processing, and technical support.

Cloud services are categorized into:

• Software as a Service (SaaS)

• Platform as a Service (PaaS)
• Infrastructure as a Service (IaaS)

Emerging types include Security as a Service (SeaaS), Knowledge as a Service, and Data
Analytics as a Service.

Major companies like Amazon, Microsoft, Google, and IBM offer a range of cloud tools and
services to benefit both users and businesses.

EMC

EMC is a leading global enterprise that needs dynamic scalability and agile infrastructure
to adapt to changing applications and business demands. Cloud computing was adopted
as the ideal solution to simplify operations and optimize infrastructure. By offering
Information Technology as a Service (ITaaS), EMC also reduces energy consumption
through resource sharing.

EMC IT

Virtualization is key to EMC IT's success, enabling on-demand resource allocation,

improved efficiency, and better resource utilization. EMC IT offers IaaS, PaaS, SaaS, and
UIaaS to its business units, providing comprehensive solutions:

1.Infrastructure as a Service (IaaS):

Allows business units to provision individual or integrated infrastructure components like

networks, storage, computing, and operating systems.

2.Platform as a Service (PaaS):

Provides secure frameworks for applications and information, offering services like Oracle
Database, SQL Server, and Greenplum as a Service, along with platforms for application
development, content management, information cycle management, security, and
integration.

3.Software as a Service (SaaS):

Delivers tools and applications, including Business Intelligence as a Service, Enterprise

Resource Planning (ERP), and Customer Relationship Management (CRM).

4.User Interface as a Service (UIaaS):

Focuses on delivering user experience rather than provisioning specific devices.

Captiva Cloud Toolkit

The Captiva Cloud Toolkit, developed by EMC, is a Software Development Kit (SDK) aimed
at helping developers integrate scanning and imaging capabilities into web-based
business applications. It is tailored for document capture vendors, commercial software
developers, and enterprises looking to create custom, scan-enabled web solutions that
complement their business offerings.
This toolkit enables developers to build scan-enabled applications in as little as one week,
reducing development, testing, and support costs. It also shortens time-to-market,
accelerates return on investment, and enhances competitiveness in the document
capture market.

Commonly Used Modules in Captiva Cloud Toolkit:

1.Scan:

This module allows importing documents into Captiva via a scanner at the page level. It
supports various formats like PDF, TIFF, and JPG. Scanning is the entry point for
documents to be processed by Captiva.

2. MDW (Multi Directory Watch)

Acts as another entry point for importing documents from folders or repositories
such as email attachments. Unlike the scan module, MDW does not interact with
scanners, making it ideal for businesses handling soft copy documents.

3.IE (Image Enhancement):

This module enhances image quality for smoother processing. Features like deskewing
and noise removal help repair unclear images. Configuration options are available to suit
specific business requirements.

4.Index:

A data capture module that extracts key information (e.g., account numbers) from
documents. It supports adding validation fields to ensure data accuracy during entry.

5.Export:

The exit point where processed images or data are sent to repositories or systems.
Exported data can be utilized for specific business needs, such as mapping captured
details to respective departments.

6.Multi:

This is the final module, used to delete batches that have successfully passed through all
other modules. It can be configured to retain batches if backups are needed.
By combining these modules, the Captiva Cloud Toolkit streamlines document processing,
reduces operational costs, and supports businesses in creating efficient and tailored
solutions for their needs.

Google
Google is a top cloud provider offering secure and reliable data storage solutions. Its
services include the Google Cloud Platform, App Engine, Cloud Print, Cloud Connect, and
more, all designed for scalability and security. Many of these services are free or available
at minimal cost, making them highly accessible and user-friendly.

Cloud Platform
Google Cloud Platform allows developers to build, test, and deploy applications on its
highly scalable and reliable infrastructure. Supported by one of the largest global
networks, it incorporates advanced software infrastructures like MapReduce, BigTable,
and Dremel for industrial innovation.

Key features include:

• Infrastructure: Virtual machines, block storage, NoSQL datastore, and big data
analytics.
• Storage: Offers versatile storage services for easy maintenance and quick data
access.
• Flexibility: Users can choose between a fully managed platform or customizable
virtual machines based on their needs.
• Integration: Simplifies integrating user applications within the cloud platform.

The platform supports automatic scaling to handle demanding workloads, scaling down
during low traffic. Designed to manage spikes in traffic, it operates like Google’s own
products. Managed services like App Engine and Cloud Datastore ensure seamless
autoscaling, and users pay only for what they use.

Cloud Storage
Google Cloud Storage is a RESTful web service for storing and accessing data on Google’s
infrastructure. It uses the REST architectural style, coordinating components, connectors,
and data elements in a distributed system. The service combines Google’s high
performance and scalability with robust security features, ensuring data is protected
through redundant storage across multiple physical locations. Google Cloud Storage is
designed to efficiently handle growing data needs while providing advanced security and
sharing capabilities, making it a reliable choice for users seeking secure and scalable
cloud storage solutions.

The following are the few tools for Google Cloud Storage:

• Google Developers Console is a web application where one can perform simple storage
management tasks on the Google Cloud Storage system.

• gsutil is a Python application that lets the user access Google Cloud Storage from the
command line.

Google Cloud Connect

Google Cloud Connect is a feature that integrates cloud services with the Microsoft Office
suite through an API. After installing a plug-in for Microsoft Office, users can save files
directly to the cloud. The cloud copy becomes the master document, accessible by
everyone, and each file is assigned a unique URL for sharing.

When changes are made to a document, those updates are reflected in real-time for all
users viewing it. If multiple people edit the same section, Cloud Connect allows users to
choose which changes to keep.

When a document is uploaded to Google Cloud Connect, the service inserts metadata into
the file. This metadata helps track changes across all copies, ensuring updates are
properly synced. The backend infrastructure is similar to the Google File System and
leverages Google Docs to sync and distribute changes to all copies using metadata.

Google Cloud Print

Google Cloud Print is a service that allows users to print from any device connected to the
Internet. To use it, the user needs a free Google profile, an app or website that supports
Google Cloud Print, and either a cloud-ready printer or a printer connected to a computer
with an active Internet connection.
When a print request is made through an app or website, it is routed via Google's servers,
which direct the request to the appropriate printer linked to the user's Google account. As
long as the printer is powered on, connected to the Internet, and has paper and ink, the
print job will be processed.

Most printers are not cloud-ready, so users often need a computer to act as a bridge.
Google Cloud Print is integrated into the Google Chrome Browser but must be explicitly
enabled. Once activated, a "connector" is used to send print commands from the
computer’s printer software to the printer.

For cloud-ready printers, direct Internet connection is possible without the need for a
computer. These printers must be registered with Google Cloud Print to be functional.
Since Google allows developers to integrate Cloud Print into their apps and websites, the
feature’s availability varies, and not all platforms support it. While Google incorporates
Cloud Print in its own products, its adoption in third-party apps and services may be
limited.

Google App Engine

Google App Engine allows users to run web applications on Google’s infrastructure,
making it easy to build, maintain, and scale applications as traffic and data storage needs
increase. With App Engine, there are no servers to manage—simply upload the app, and
it’s ready to serve users. The app can be hosted on a custom domain (e.g.,
www.example.com) Or a free appspot.com domain. Access can be set to public or
restricted to an organization.

App Engine supports multiple programming languages, including Java, Python, PHP, and
Go. It provides environments tailored for each language, such as Java with the JVM and
servlets, Python with a fast interpreter, and go with native support for compiled code. App
Engine’s infrastructure ensures that applications run quickly, securely, and without
interference from others.

Users pay only for the resources they use, including storage and bandwidth, which are
measured in gigabytes and billed competitively. App Engine offers a free tier, providing up
to 1 GB of storage and enough resources to support about five million page views per
month. Once billing is enabled, users can scale beyond the free limits and only pay for
additional resources consumed.
Amazon Web Services (AWS)
Amazon Web Services (AWS) is a suite of remote computing services, collectively forming
a cloud computing platform provided over the Internet by Amazon.com. Key services
include Amazon Elastic Compute Cloud (EC2), Amazon Simple Queue Service (SQS), and
Amazon S3. Amazon EC2 is a computing service, while SQS and S3 are support services.
EC2 offers scalable computing capacity, allowing businesses to deploy virtual servers
much faster and cheaper than building physical server farms. AWS data centers are spread
globally across locations like Ashburn, Virginia, Dallas/Fort Worth, Los Angeles, Miami,
Newark, New Jersey, Palo Alto, California, Seattle, St. Louis, Amsterdam, Dublin,
Frankfurt, London, Hong Kong, Singapore, and Tokyo. AWS provides a highly reliable and
cost-efficient cloud infrastructure for various business needs.

Amazon Elastic Compute Cloud (EC2)

Amazon EC2 is an Infrastructure as a Service (IaaS) offered by AWS, making it one of the
leading IaaS providers. Powered by Amazon's extensive infrastructure, originally built for its
retail business, EC2 provides a virtual computing environment with various instance types,
operating systems, and software packages. The platform allows users to instantiate virtual
machines (VMs) and adjust their capacity and characteristics through web service
interfaces, hence the term "elastic."

EC2 instances are created by booting Amazon Machine Images (AMIs), which contain all
the necessary information to launch an instance. Users can interact with EC2 using the
AWS Management Console (GUI) or a web service API that supports protocols like SOAP
and Query Requests. The API provides libraries and resources for multiple programming
languages including Java, PHP, Python, Ruby, Windows, and .NET. The virtual
infrastructure is powered by the Xen hypervisor.

Key Features of EC2:

• Instance Types:
❖ Standard instances: Suitable for general applications.
❖ Micro instances: For low-throughput applications.
❖ High-memory instances: Ideal for high-throughput applications.
❖ High-CPU instances: For compute-intensive tasks.
❖ Cluster compute instances: Best for high-performance computing (HPC).
• On-Demand & Reserved Instances:
Instances are billed hourly, and users can reserve instances at discounted rates.
Spot Instances can be obtained by bidding on unused capacity.
• Regions and Availability Zones:
Instances can be placed across multiple locations defined by regions and
availability zones. This ensures fault tolerance and reliability.
 • Monitoring and Management:
Instances can be monitored and managed through the AWS Management Console
or API. Amazon CloudWatch provides visibility into resource utilization, operational
performance, and demand patterns (e.g., CPU utilization, disk I/O, and network
traffic).
• Security:
EC2 instances are authenticated using a signature-based protocol with key pairs.
Amazon VPC allows integration with existing IT infrastructure via a virtual private
network (VPN), offering isolated computing resources with customizable security
settings like firewalls and intrusion detection.
• Elastic Load Balancing (ELB):
ELB distributes incoming application traffic across multiple instances based on
metrics such as request count and latency. It ensures fault tolerance and
automatic scaling of resources.
• Persistent Storage:
EC2 instances can be paired with storage solutions like Elastic Block Storage (EBS)
for secure, reliable, and scalable storage or Amazon S3 for highly durable storage of
mission-critical data. EBS volumes can be attached to instances as block devices,
while S3 allows data storage as objects.

Amazon EC2 is a robust and flexible platform for running applications on the cloud, with
capabilities that ensure high availability, scalability, and security.

Amazon Simple Storage Service (S3)

Amazon S3 is a web-scale storage service designed to make computing easier for
developers. It offers a simple web service interface to store and retrieve an unlimited
amount of data at anytime from anywhere on the web. S3 allows developers to use the
same scalable, reliable, and secure infrastructure that Amazon employs to run its global
network of websites. By providing these benefits, Amazon S3 offers developers access to
cost-effective and powerful storage solutions.

S3 is designed to be highly scalable, reliable, fast, and inexpensive, making it an ideal

solution for a variety of storage needs. It ensures high durability and availability of data,
making it suitable for storing content such as web applications, media files, backup data,
and large datasets. With its pay-as-you-go pricing model, users only pay for the storage
they use, maximizing cost efficiency.
Key Features of Amazon S3:

• Scalability and Reliability:

S3 is highly scalable, allowing users to store large amounts of data and handle fluctuating
data loads with ease. Its reliability ensures that data is always available when needed.

• Reduced Redundancy Storage (RRS):

S3 provides a cost-effective solution for storing easily reproducible content or data where
an original copy is stored elsewhere. RRS reduces storage costs by providing lower
redundancy for less critical data.

• Data Analytics and Computation:

S3 is well-suited for storing data for analysis and computation. Data stored in S3 can be
transferred to Amazon EC2 for large-scale analytics or computations without incurring
data transfer costs between services.

• Backup and Archiving:

S3 provides a secure, durable solution for backup and archiving of critical data. The AWS
Import/Export feature enables users to move large amounts of data into and out of AWS
using physical storage devices, ideal for disaster recovery or periodic backups.

• Static Website Hosting:

S3 offers static website hosting for websites that serve static content, such as HTML files,
images, videos, and JavaScript. It’s a simple and cost-effective way to host content without
the need for a web server.

• Performance and Security:

Amazon S3 is built with performance and security in mind. Data is stored securely and can
be accessed via a highly reliable web interface. It also supports encryption to ensure data
protection.

Amazon S3 is an essential tool for developers and businesses looking to store and manage
large volumes of data, providing flexibility, scalability, and security, along with an easy-to-
use interface.
Amazon Simple Queue Service (SQS)
Amazon SQS is a fast, reliable, scalable, and fully managed message queuing service
offered by AWS. It simplifies the decoupling of the components within a cloud application,
making it cost-effective to transmit data at any volume and throughput. The service
ensures that messages are not lost and does not require other services to always be
available, providing reliability for applications that depend on message-driven
architecture.

As a distributed queue system, Amazon SQS enables web service applications to quickly
and reliably queue messages. These messages can be generated by one component and
consumed by another, making it easy to handle communication between different parts of
an application. A queue serves as a temporary repository for messages awaiting
processing.

Key Features of Amazon SQS:

• High Availability and Scalability:

Amazon SQS is built on a redundant infrastructure that ensures high availability for sending
and receiving messages. It can scale easily to handle varying volumes of messages.

• Multiple Readers and Writers:

SQS supports multiple readers and writers, allowing different components of an

application to interact with the queue simultaneously, which enhances performance and
flexibility.

• Access Control and Security:

The service offers built-in access control facilities, ensuring secure interactions with the
queue. Users can define who can send and receive messages, adding an extra layer of
security.

• Variable Length Messages:

SQS allows for variable-length messages, providing flexibility in the size of the data that
can be transmitted.
 • Configurable Queue Settings:

Users can customize the settings of each queue to meet their specific needs, such as
adjusting message retention times and setting visibility timeouts.

• Message Delivery Guarantees:

SQS guarantees the reliable delivery of messages, even during high traffic volumes or if
there are failures in one part of the system.

Amazon SQS is an essential tool for building distributed, message-driven applications that
require high availability, scalability, and performance without the complexity of managing
message queues manually. It supports a wide range of use cases from simple message
delivery to complex event-driven architectures.

Microsoft
Microsoft IT (MSIT) has made cloud computing the preferred environment for new and
migrated applications. MSIT developed a methodology and best practices to assess and
migrate applications to the cloud, ensuring optimal performance and cost-efficiency.
These insights are shared with other organizations to help them transition smoothly to
cloud computing. By embracing the cloud, Microsoft empowers businesses to achieve
scalability, flexibility, and cost savings while maintaining performance.

Additionally, Microsoft provides tools like Azure to support businesses in their digital
transformation. Azure offers a comprehensive set of cloud services, including compute,
storage, and networking, helping companies build, deploy, and manage applications. With
a focus on security, compliance, and innovation, Microsoft ensures that organizations can
leverage cloud technologies effectively and safely. As the cloud landscape evolves,
Microsoft continues to support businesses with cutting-edge solutions for long-term
growth.

Windows Azure
Windows Azure Cloud Services (web and worker roles/PaaS) offer developers an easy way
to deploy and manage application services while Windows Azure handles the underlying
infrastructure and operating system.
The Migration Assessment Tool (MAT) helps determine the effort required to migrate an
application to Windows Azure by generating a report based on responses to binary
questions. The Windows Azure Pricing Calculator compares an application’s current
operational costs with potential costs on Windows Azure and SQL Azure, helping
businesses evaluate cost-effectiveness.

The Windows Azure Pack for Windows Server provides a collection of Windows Azure
technologies that allow Microsoft customers to create a self-service, multitenant cloud in
their data centers, running on Windows Server 2012 R2 and System Center 2012 R2. This
technology offers an experience similar to the public Windows Azure, enabling businesses
to enhance their infrastructure.

Microsoft Assessment and Planning Toolkit (MAP)

The Microsoft Assessment and Planning Toolkit (MAP) is an agentless, automated tool
designed for cloud migration planning and assessment. MAP generates detailed readiness
reports, executive proposals, and provides hardware and software information. It also
offers recommendations to accelerate the application migration process for both private
and public cloud environments. MAP analyzes server utilization data for virtualization and
consolidation with Hyper-V, helping organizations streamline their cloud transition
process.

SharePoint
Microsoft SharePoint is a web application platform offering a wide range of collaboration
tools backed by a common technical infrastructure. Integrated with the Microsoft Office
suite, SharePoint provides a user-friendly interface that is accessible to nontechnical
users. It is designed to enable intranet portals, document management, collaboration,
enterprise search, and business intelligence. Additionally, SharePoint offers system
integration, process integration, and workflow automation capabilities. While not a free
tool like Google Cloud Connect, SharePoint’s advanced features set it apart from other
collaboration platforms.

IBM Cloud
IBM is a prominent player in cloud computing, offering a range of solutions through its IBM
SmartCloud. The platform includes IaaS, SaaS, and PaaS, available via public, private, and
hybrid cloud delivery models. IBM’s cloud offerings begin with physical hardware
supporting virtualization, followed by IBM Websphere for application infrastructure. The
management layer uses IBM Tivoli middleware, providing tools for automated provisioning,
monitoring, and usage tracking. Integrated workload tools support cloud-based
collaboration, development, analytics, and security, helping businesses optimize their
cloud operations effectively.

Cloud Models
IBM provides a range of cloud models to suit different business needs. These include fully
private clouds, where companies manage their infrastructure, as well as hybrid clouds,
which combine public and private elements. The flexibility in choosing between virtual
private cloud services and public cloud services allows businesses to scale operations
securely and efficiently. IBM also offers secure private cloud solutions for businesses
concerned about security risks, with tools like IBM Workload Deployer and Cloudburst to
facilitate the building and deployment of private clouds. These customizable options give
enterprises the ability to integrate and manage their workloads in a way that best fits their
requirements.

IBM SmartCloud
IBM SmartCloud is a comprehensive suite of cloud computing products and solutions,
encompassing IaaS, PaaS, SaaS, and BPaaS across public, private, and hybrid cloud
models. It is organized into three key areas: SmartCloud Foundation, SmartCloud
Services, and SmartCloud Solutions. SmartCloud Foundation provides the infrastructure,
hardware, provisioning, and management for private or hybrid clouds. Built on this
foundation, SmartCloud Services includes IaaS, PaaS, and backup services, while
SmartCloud Solutions offers a range of SaaS applications for collaboration, analytics, and
marketing. BPaaS (Business Process as a Service) extends the cloud offering by delivering
business processes through the cloud with elastic scaling and web interfaces.
SAP Labs
SAP Labs is a leading developer of enterprise software that helps businesses manage
operations and customer relations. Known for its dominance in the enterprise applications
market, SAP’s best-known products include SAP ERP (Enterprise Resource Planning), SAP
BW (Business Warehouse), SAP Business Objects, and SAP HANA, an in-memory
computing appliance. Additionally, SAP has expanded its portfolio with Sybase mobile
products. As one of the largest software companies globally, SAP continues to innovate in
the areas of business management and data processing.

SAP HANA Cloud Platform

SAP HANA Cloud Platform is a modular Platform-as-a-Service (PaaS) that supports open
standards and is Eclipse-based. It allows developers to deploy applications using
command-line tools, either as web application archive (WAR) files or OSGi bundles, within
a Java-based runtime environment powered by SAP HANA. Key features of SAP HANA
Cloud Platform include:

• Enterprise platform built for developers

• Native integration with SAP and non-SAP software
 • In-memory persistence
• Secure data platform
• Lightweight, modular runtime container for applications

This platform enables rapid development and deployment of business and consumer
applications, offering seamless integration with minimal coding requirements, and
supports open programming standards to enhance developer productivity.

Virtualization Services Provided by SAP

SAP provides virtualization services to enhance the efficiency and performance of ERP
applications. Virtualizing ERP applications offers several business benefits, including:

• Maximizing hardware utilization: Ensures optimal use of physical resources,

leading to better performance.
• Shorter development cycles: Accelerates project timelines by streamlining
application deployment.
• Reduction in IT costs: Minimizes the need for physical hardware and reduces
infrastructure expenses.
• Improved availability: Increases system uptime and reliability by leveraging virtual
environments.
• Energy savings: Reduces power consumption through efficient hardware
utilization.

A joint service from SAP and VMware supports the transition to a more open and flexible
private cloud platform, leveraging proven virtualization technology to deliver these
benefits.

Salesforce
Salesforce.com is a cloud computing and social enterprise SaaS provider based in San
Francisco. Of its cloud platforms and applications, the company is best known for its
Salesforce CRM product, which is composed of Sales Cloud, Service Cloud, Marketing
Cloud, Force.com, Chatter, and Work.com. In addition to its products and platforms,
Salesforce.com created AppExchange, a custom application building and sharing
platform. The company also has consulting, deployment, and training services.
Sales Cloud
Sales Cloud is a module in Salesforce.com designed to streamline the entire sales
process. It includes features such as Leads, Accounts, Contacts, Contracts,
Opportunities, Products, Pricebooks, Quotes, and Campaigns, with limits applying in
certain cases. It offers functionalities like web-to-lead for online lead capture and
autoresponse rules. Sales Cloud manages contact information and integrates social
media, real-time customer collaboration, and Chatter to enhance customer engagement.
This platform provides up-to-date account information and social insights, all available in
one place, anytime, anywhere. Sales Cloud ensures real-time updates, from contact
details to deal changes and discount approvals. Salesforce built it to be user-friendly,
similar to consumer websites like Amazon, and in the cloud to eliminate the risks and
costs of traditional software. With its open architecture and automatic updates, Sales
Cloud removes the hidden costs and lengthy implementations associated with traditional
CRM software, continuously innovating with mobile, collaboration, and social intelligence
to stay ahead of the competition.

Service Cloud: Knowledge as a Service

Service Cloud is the customer service module in Salesforce.com, offering a range of
features like Accounts, Contacts, Cases, and Solutions. It includes functionalities such as
a public knowledge base, web-to-case, call center integration, and a self-service portal,
along with customer service automation. Service Cloud introduces the world’s first
enterprise-grade knowledge base running entirely on a multitenant cloud platform,
eliminating the need for expensive data centers or on-premises software. This knowledge
base is fully integrated with the entire CRM solution, enabling users to manage knowledge
as a process—creating, reviewing, delivering, analyzing, and improving it. Service Cloud
ensures that agents have the right answers readily available, whether through phone,
email, or chat, while also allowing knowledge to be shared via the company’s website or
social channels like Twitter and Facebook. It combines enterprise-scale knowledge
management with the ease of use that Salesforce is known for, making it suitable for
businesses of any size or complexity.
Rackspace
Rackspace Cloud, a division of Rackspace, offers cloud computing services with a focus
on Infrastructure as a Service (IaaS). It provides three main solutions: Cloud Servers, Cloud
Files, and Cloud Sites. Cloud Servers delivers on-demand computational power with
virtual machines configured with varying capacities, available in different flavors and
images. These virtual machines are virtualized using Xen Hypervisor for Linux and Xen
Server for Windows, with sizes ranging from 256 MB to 15.5 GB of memory. Users can
create custom images for flexibility, and backup schedules can be set to ensure data
recovery in case of failures.

Cloud Servers are managed through the Rackspace Cloud Control Panel (GUI) or via API
using a RESTful interface, allowing users to automate scaling and load balancing. Servers
are equipped with RAID10 disk storage for persistent data storage, and Cloud Servers can
scale automatically to balance loads. Rackspace provides support through a knowledge
base, forums, and live chat, while the API is open-sourced and supports high-level
languages like C++, Java, Python, and Ruby. The platform also offers the option of a load
balancing solution, currently in beta, and provides additional processing power at no extra
cost when available.

VMware
VMware is a leader in virtualization technology and provides enterprise cloud computing
solutions for both private and public clouds, as well as hybrid cloud environments.
VMware's suite of products, including VMware vCloud Director, VMware vCloud
Datacenter Services, VMware vSphere, and VMware vShield, help organizations optimize IT
infrastructure and leverage cloud services efficiently. VMware’s private cloud offerings
enhance operational efficiency by consolidating physical infrastructures, providing greater
standardization, rapid provisioning, and self-service. These solutions lead to cost savings
and improved resource management for enterprises.

VMware vSphere transforms IT infrastructures into virtual storage, compute, and network
resources, offering both infrastructure and application-level services. At the infrastructure
level, it facilitates the operation and management of resources, while at the application
level, it ensures secure, scalable, and available applications. VMware vCloud Director,
used with VMware vSphere, enables enterprises to build secure, multitenant private
clouds by pooling resources into virtual datacenters, which can be accessed through
automated, catalog-based services. VMware vShield provides security services like
perimeter protection, firewalls, and VPN to safeguard these environments.

VMware also offers public and hybrid cloud solutions through partnerships with certified
service providers, enabling users to extend their private cloud into a public cloud
environment. VMware vCloud Datacenter Services and VMware vCloud Express provide
scalable, reliable, and pay-as-you-go infrastructure, allowing users to augment their
internal resources with external cloud resources. VMware’s vCloud API enables
automated orchestration of operational tasks and workflows, further enhancing the
flexibility and efficiency of VMware’s cloud solutions.

Manjrasoft
Manjrasoft is a lesser-known provider of cloud services, but it offers a platform called
Aneka that simplifies the development of applications. Aneka provides a set of services
designed to accelerate the development, deployment, and management of applications,
helping businesses save time and money. The platform is market-oriented and aims to
enhance business productivity and profitability by providing efficient cloud computing
solutions tailored to meet business needs. Manjrasoft's cloud services focus on building
scalable and high-performance applications, ultimately driving business growth.

Aneka Platform
The Aneka platform simplifies enterprise cloud construction and application development
while maintaining flexibility, scalability, reliability, and extensibility. Key features of Aneka
include:

1. Configurable and Flexible Execution Platform: Aneka offers a container enabling

pluggable services and security implementations, supporting multiple
authentication/authorization mechanisms such as role-based security and
Windows domain–based authentication.
2. Multiple Persistence Options: The platform supports various persistence options,
including Relational Database Management Systems (RDBMS), SQL Express,
MySQL, and flat files.
 3. Software Development Kit (SDK): Aneka provides an SDK supporting multiple
programming models, such as the object-oriented thread model, task model for
legacy applications, and the MapReduce model for data-intensive applications.
4. Custom Tools: Tools like the Design Explorer are available for parameter sweep
studies, enhancing the flexibility of development.
5. Management Tool: Aneka includes an easy-to-use management tool for Service
Level Agreement (SLA) and Quality of Service (QoS) negotiation, as well as dynamic
resource allocation.
6. Cloud Deployment: It supports the deployment of applications on both private and
public clouds, with seamless integration between them.

Aneka enables servers and desktop PCs to be linked together, forming a powerful
computing infrastructure. This approach allows businesses to be more energy-efficient
and cost-effective without the need for extensive hardware investments. Each Aneka node
includes configurable containers for information indexing, scheduling, execution, and
storage services, supporting various programming models, security protocols, persistence
options, and communication protocols.
Summary
This chapter covered various companies offering cloud computing tools and technologies
to help users adapt to the cloud environment. It highlighted the cloud features provided by
each company, with some services like Google Docs and Google Cloud Print being free,
while others, like AWS and Microsoft, are proprietary.A summary of tools and services
offered by each company is provided in Table 11.1, and Table 11.2 outlines pricing, service
models, and deployment models for selected providers. Additional details can be found in
the references section.