Module- 3

CLOUD PLATFORM ARCHITECTURE

1. CLOUD COMPUTING AND SERVICE MODELS:

Ø Over the past two decades, the world economy has rapidly moved from manufacturing
to more service-oriented.
Ø Cloud computing benefits the service industry most and advances business computing
with a new paradigm.
Ø Developers of innovative cloud applications no longer acquire large capital equipment in
advance. They just rent the resources from some large datacenters that have been
automated for this purpose.

1.1 PUBLIC CLOUD:

Ø A public cloud is built over the Internet and can be accessed by any user who has paid
for the service. Public clouds are owned by service providers and are accessible through
a subscription.
Ø The providers of the aforementioned clouds are commercial providers that offer a
publicly accessible remote interface for creating and managing VM instances within
their proprietary infrastructure.
Ø A public cloud delivers a selected set of business processes. The application and
infrastructure services are offered on a flexible price-per-use basis.
Examples:
• Google App Engine (GAE)
• Amazon Web Services (AWS)
• Microsoft Azure
• IBM Blue Cloud
• [Link]’s [Link].
Advantages:
• Standardization
• Preserves Capital Investment
• Offers Application Flexibility

1.2 PRIVATE CLOUD:

Ø A private cloud is built within the domain of an intranet owned by a single organization.
It is client owned and managed, and its access is limited to the owning clients and their
partners.
 Ø Its deployment was not meant to sell capacity over the Internet through publicly
accessible interfaces.
Ø Private clouds give local users a flexible and agile private infrastructure to run service
workloads within their administrative domains.
Ø A private cloud is supposed to deliver more efficient and convenient cloud services. It
may impact the cloud standardization, while retaining greater customization and
organizational control.
Examples:
• IBM RC2
• Amazon Virtual Private Cloud
• VMware Private Cloud
• Rackspace Private Cloud (Powered by OpenStack)
• CloudBees

Advantages:
• Customization & offers higher efficiency
• Resiliency
• Security
• Privacy

1.3 HYBRID CLOUDS:

Ø A hybrid cloud is built with both public and private clouds Private clouds can also support
a hybrid cloud model by supplementing local infrastructure with computing capacity from
an external public cloud.
 Ø A hybrid cloud provides access to clients, the partner network, and third parties. Hybrid
clouds operate in the middle, with many compromises in terms of resource sharing.

Example:
Ø Research Compute Cloud (RC2) is a private cloud, built by IBM, that interconnects the
computing and IT resources at eight IBM Research Centers scattered throughout the
United States, Europe, and Asia.

1.4 CLOUD SERVICES:

1.4.1 Infrastructure as a Service(IaaS):

Ø This model allows users to use virtualized IT resources for computing, storage, and
networking. The service is performed by rented cloud infrastructure.
Ø The user can deploy and run his applications over his chosen OS environment. The user
does not manage or control the underlying cloud infrastructure, but has control over the
OS, storage, deployed applications, and possibly select networking components.
Ø This IaaS model encompasses storage as a service, compute instances as a service, and
communication as a service.
Ø Many startup cloud providers have appeared in recent years. GoGrid, FlexiScale, and
Aneka are good examples.

1.4.2 .Platform as a Service (PaaS):

Ø To be able to develop, deploy, and manage the execution of applications using

provisioned resources demands a cloud platform with the proper software
environment.
 Ø Such a platform includes operating system and runtime library support. This has
triggered the creation of the PaaS model to enable users to develop and deploy their
user applications.
Ø The platform cloud is an integrated computer system consisting of both hardware and
software infrastructure.
Ø The user application can be developed on this virtualized cloud platform using some
programming languages and software tools supported by the provider (e.g., Java,
Python, .NET). The user does not manage the underlying cloud infrastructure. The cloud
provider supports user application

1.4.3 Software as a Service (SaaS):

Ø This refers to browser-initiated application software over thousands of cloud customers.

Services and tools offered by PaaS are utilized in construction of applications and
management of their deployment on resources offered by IaaS providers.
Ø The SaaS model provides software applications as a service. As a result, on the customer
side, there is no upfront investment in servers or software licensing.
Ø On the provider side, costs are kept rather low, compared with conventional hosting of
user applications. Customer data is stored in the cloud that is either vendor proprietary
or publicly hosted to support PaaS and IaaS.
Examples of SaaS:
• Google Gmail and docs
• Microsoft SharePointa
• CRM software from [Link].

5. Centralized vs. Distributed Cloud Computing

There are two contrasting views on how cloud computing operates. Some believe it is centralized
computing because data is stored and managed in large data centers owned by cloud providers. Others
argue that it is distributed computing since workloads are processed across multiple virtual machines
spread across different locations. In reality, cloud computing is a distributed system that appears
centralized to users. For example, Microsoft’s Chicago data center houses 100,000 servers, but its
cloud services are accessed worldwide, demonstrating a distributed architecture.
6. Cloud Networking and Data Centers

Cloud computing relies on large data centers connected through high-speed networks to deliver
services. A cloud platform typically has three types of nodes: compute nodes for processing tasks,
control nodes for managing cloud operations, and gateway nodes for securing access. Data centers are
optimized to handle fluctuating workloads efficiently. Google’s data centers, for example, process
massive amounts of data for services like YouTube and Google Search. The network structure of data
centers ensures fast access, high availability, and secure storage of data.

Data-Center Design and Interconnection Networks

1. Introduction to Data-Center Design

A data center is a facility that houses a large number of servers connected through a high-speed
interconnection network. These servers process, store, and manage vast amounts of data for cloud
computing, web applications, and enterprise services. Data centers vary in size, from small modular
units to warehouse-scale facilities with hundreds of thousands of servers. The efficiency of a data
center depends on its design, cooling system, interconnection network, and fault tolerance
mechanisms. The evolution of data-center technology has led to improved scalability, cost-
effectiveness, and energy efficiency.

2. Warehouse-Scale Data-Center Design

Warehouse-scale data centers are massive facilities, often as large as shopping malls, with the capacity
to house 400,000 to 1 million servers. These large data centers reduce operational costs by leveraging
economies of scale—as the number of servers increases, the cost per unit decreases. Smaller data
centers, such as those with 1,000 servers, have higher costs per unit due to network and storage
expenses. Large-scale data centers, like those operated by Microsoft, Amazon, and Google, are
strategically distributed across the globe to enhance performance and reliability.
3. Data-Center Construction Requirements

Data centers are built using commercially available hardware components. A typical server includes
multi-core processors, DRAM (memory), and disk storage. Servers are grouped in racks, with each
rack connected to a rack-level switch for internal communication. The racks are further connected to a
cluster-level switch that manages communication across the entire data center. Data storage and
bandwidth distribution are critical considerations, as applications must handle differences in latency,
bandwidth, and storage capacity. Large-scale data centers use redundant hardware and software
systems to ensure reliability and prevent data loss during failures.

4. Cooling System in Data Centers

Efficient cooling is essential to prevent server overheating and maintain optimal performance. Data
centers use a raised floor system, which allows for the distribution of cool air from a cooling unit
(CRAC - Computer Room Air Conditioning). Servers are arranged in alternating cold and hot aisles
to prevent the mixing of hot and cold air. Newer data centers utilize water-based cooling towers that
dissipate heat more effectively. The cooling system ensures energy efficiency, reduces hardware
failures,
208 andCHAPTER
lowers operational
4 Cloudcosts.
Platform Architecture over Virtualized Data Centers

Ceiling Ceiling

Liquid supply Rack Rack Rack Rack Liquid supply

CRAC CRAC
unit unit

Floor tiles Floor tiles

Floor slab Floor slab

FIGURE 4.9
The cooling system in a raised-floor data center with hot-cold air circulation supporting water heat exchange
facilities.
(Courtesy of DLB Associates, D. Dyer [22] )

about 2–4 ft above the concrete floor. The under-floor area is often used to route power cables to
racks, but its primary use is to distribute cool air to the server rack. The CRAC (computer room air
conditioning) unit pressurizes the raised floor plenum by blowing cold air into the plenum.
5. Data-Center
The cold Interconnection
air escapes from Networks
the plenum through perforated tiles that are placed in front of server
racks. Racks are arranged in long aisles that alternate between cold aisles and hot aisles to avoid
mixing hot
A critical and cold of
component [Link]-center
The hot air design
producedis by
thethe servers circulatesnetwork,
interconnection back to the intakes
which of the
links servers for
CRAC units that cool it and then exhaust the cool air into the raised floor plenum again. Typically,
efficient data transfer. This network must meet five key requirements:
the incoming coolant is at 12–14°C and the warm coolant returns to a chiller. Newer data centers
often insert a cooling tower to pre-cool the condenser water loop fluid. Water-based free cooling
usesLow latency
cooling (fast
towers to data transfer)
dissipate heat. The cooling towers use a separate cooling loop in which
water absorbs the coolant’s heat in a heat exchanger.
High bandwidth (large data capacity)
4.2.2 Data-Center Interconnection Networks
A critical core design of a data center is the interconnection network among all servers in the data-
center cluster. This network design must meet five special requirements: low latency, high band-
 Low cost

Message-passing interface (MPI) communication support

Fault tolerance

The network design must support both point-to-point communication (direct server connections) and
collective communication (group data transfers). This ensures smooth operations for cloud computing,
big data processing, and distributed applications.

6. Network Expandability in Data Centers

As data centers grow, the network infrastructure must scale efficiently. Modern data centers use
modular designs, where entire server containers (housing hundreds of servers) can be added
seamlessly. This approach reduces costs and simplifies maintenance. Load balancing mechanisms
ensure that no single network link becomes a bottleneck, preventing slowdowns in data processing. Fat-
tree and crossbar network topologies are commonly used to maintain high performance and prevent
congestion.

7. Fault Tolerance and Graceful Degradation

In a large-scale data center, hardware or software failures are inevitable. The interconnection network
must be designed to handle failures without affecting overall performance. This is achieved through
redundancy, where multiple paths exist between servers to ensure continuous operation. Hot-
swappable components (such as power supplies and switches) allow for easy replacement without
downtime. Graceful degradation ensures that if a failure occurs, the system continues running with
minimal disruption. Software solutions, such as automated rerouting and error detection, further
enhance network reliability.

8. Switch-Centric vs. Server-Centric Network Design

Data-center networks are built using either switch-centric or server-centric architectures:

Switch-centric networks rely on network switches to manage connections between servers. This is
simpler to implement but depends heavily on switch performance.

Server-centric networks modify the server software to help manage traffic, improving efficiency but
requiring specialized configurations.

Most modern data centers prefer a switch-centric approach due to its scalability and ease of
deployment.

1. Modular Data Centers in Shipping Containers

A modular data center is structured as a collection of shipping container-sized units, each containing
server clusters. These containers are designed to be mobile, truck-towed, and easily deployed in
different locations. Inside each container, hundreds of blade servers are mounted in racks along the
walls. A specialized cooling system uses fans to push hot air through a heat exchanger, cooling it before
recirculating. The SGI ICE Cube modular data center, for example, can house 46,080 processing cores
or provide up to 30 PB of storage in a single container. These modular designs offer energy efficiency,
reducing cooling costs by up to 80% compared to traditional warehouse-style data centers. Additionally,
they can be moved to locations with lower power costs, better cooling water availability, and more
affordable housing for maintenance staff.

2. Construction of Container Data Centers

Containerized data centers consist of a network, computing infrastructure, storage, and cooling
mechanisms all integrated within a weatherproof and transportable structure. Cooling efficiency is
enhanced by optimizing airflow and water circulation, which helps regulate the temperature inside.
Since seasonal changes affect cooling requirements, the system must be adaptable. The construction
process follows a staged approach—starting with individual servers, then racks, and finally entire
container units. For example, assembling a rack of 40 servers takes about half a day, whereas
constructing a full container-based system with 1,000 servers requires careful planning for floor space,
power supply, networking, and testing. The entire setup can be completed within a few days if all
components and utilities are available. These modular data centers are particularly useful for industries
like healthcare, where localized data centers at multiple clinic sites can improve service delivery.
However, ensuring smooth data synchronization and periodic updates between local and central
databases remains a key challenge.

3. Interconnection of Modular Data Centers

To create large-scale cloud infrastructure, multiple modular data centers must be interconnected. Inside
each container, the BCube network topology is commonly used to connect individual servers. However,
when multiple containers need to be connected, an additional layer of networking, called MDCube, is
introduced. The MDCube network architecture links multiple BCube containers through high-speed
switches, forming a large, scalable cloud infrastructure. This system builds a virtual hypercube at the
container level, in addition to the cube structure inside each container. The MDCube interconnection
enables efficient communication between modular data centers, supporting cloud applications that
require high-speed data exchange. Various configurations of MDCube networks can be used to optimize
performance, scalability, and fault tolerance.

4. Data-Center Management Issues

Managing a modular data center involves several key considerations. First, user satisfaction is essential,
meaning that data centers should be designed to provide stable and high-quality service for at least 30
years. Second, information flow must be controlled to ensure smooth data access, high availability, and
uninterrupted services. Third, multi-user management is necessary to support various functions, such as
database updates, traffic control, and server maintenance. Scalability is another crucial factor, as data
centers must be able to expand as workloads grow. This requires adaptable storage, computing power,
cooling, and power supply systems. Reliability is also critical, with failover mechanisms, fault tolerance,
and live virtual machine migration integrated to prevent downtime in case of system failures. Cost
efficiency is equally important, as data centers should minimize operational expenses while delivering
high performance to users and cloud service providers.

5. Security and Sustainability in Data Centers

Security in modular data centers focuses on data privacy, protection from cyber threats, and maintaining
system integrity. Strong defense mechanisms are required to prevent unauthorized access, network
attacks, and data breaches. Additionally, green IT practices are becoming a priority in data center
design. Reducing power consumption, optimizing energy efficiency, and using advanced cooling
techniques help minimize the environmental impact. Modern data centers implement smart energy
management to reduce electricity costs and carbon footprints while maintaining high-performance
computing capabilities.

6. Marketplaces in Cloud Computing Services

Container-based data centers are increasingly being deployed using factory-assembled racking,
stacking, and packing methods. This approach reduces complexity at the installation site and improves
efficiency. Unlike traditional data centers, which are often custom-built, modular units allow for a more
standardized and scalable infrastructure. These containerized data centers achieve high power densities,
exceeding 1250 W per square foot, making them space-efficient. They can be installed in various
locations, including rooftops and parking lots, offering flexibility in deployment. Moreover, sufficient
redundancy must be built into the system to allow future upgrades without requiring major overhauls.