Cloud Computing

Cloud computing is a service delivery model that is often built on distributed computing technology.

Cloud deployment models define where the cloud infrastructure is located and who controls and uses it. There are four main models.

1. Public cloud

Infrastructure is owned by a provider and offered to the general public over the Internet. Examples: AWS, Azure, Google Cloud.

• Benefits: low cost, pay-per-use, no maintenance, near-unlimited scalability, quick provisioning.
• Limitations: less control and customisation, security and compliance concerns, shared resources, possible vendor lock-in.

2. Private cloud

Infrastructure is operated for a single organisation, on premises or hosted by a third party.

• Benefits: greater security, control, privacy, compliance and customisation, predictable performance.
• Limitations: high capital and maintenance cost, limited scalability, needs skilled staff.

3. Community cloud

Infrastructure is shared by several organisations with common concerns (mission, security, compliance), for example universities or government agencies.

• Benefits: cost shared among members, better security than public, collaboration.
• Limitations: limited capacity, shared responsibility and governance disputes, higher cost than public.

4. Hybrid cloud

A composition of two or more of the above, linked by standard technology.

• Benefits: flexibility, cloud bursting, optimised cost and security by workload placement.
• Limitations: complex integration and management, security consistency, networking latency.

Choosing a model for an application: use public cloud for web apps and testing with variable load, private cloud for sensitive data, community cloud for shared sector needs, and hybrid when sensitive data must stay private while the application scales on public resources.

Cloud computing is a model for delivering computing resources (servers, storage, databases, networking, software and analytics) over the Internet on demand, on a pay-per-use basis, with minimal management effort. NIST defines it as a model for enabling convenient, on-demand network access to a shared pool of configurable computing resources that can be rapidly provisioned and released.

Characteristics

1. On-demand self-service: users provision resources themselves without human interaction with the provider.
2. Broad network access: services are available over the network from any device.
3. Resource pooling: the provider's resources serve many tenants using a multi-tenant model.
4. Rapid elasticity: resources scale up or down quickly to match demand.
5. Measured service: usage is metered, giving pay-as-you-go billing and transparency.
6. Other features: high availability, reliability, cost reduction and automatic updates.

OS-level virtualization creates isolated user-space instances (containers) that share one host operating-system kernel. Examples are Docker, LXC and OpenVZ.

Advantages

1. Low overhead and high performance: there is no separate guest OS or hypervisor layer, so performance is near native.
2. Fast start-up: containers start in seconds.
3. Higher density: many more instances run on one machine because they share the kernel and use less memory and storage.
4. Efficient resource use: CPU, memory and I/O are allocated and limited flexibly.
5. Easy portability and deployment: an application with its dependencies runs the same everywhere.
6. Simple management: one OS to patch and maintain, and quick cloning and scaling.
7. Cost saving: less hardware and fewer licences are needed.
8. Isolation: processes, file systems and networks of different containers are kept separate.

MapReduce is a programming model that processes large data sets in parallel across a cluster, using a map step and a reduce step.

Applications

1. Distributed word count and text processing: counting word frequencies in large corpora.
2. Distributed grep and log analysis: searching for patterns in huge log files and analysing web server logs.
3. Web indexing and search: building inverted indexes for search engines (Google's original use).
4. Web link analysis: computing PageRank and reverse web-link graphs.
5. Machine learning and data mining: clustering, classification and recommendation on big data.
6. Sorting and large-scale joins: processing database-style data.
7. Scientific and bioinformatics computing: genome sequence analysis and astronomy data.
8. Business analytics: sales analysis, fraud detection and ad click analysis.

SaaS security is the set of controls that protect applications and data delivered as a service over the Internet. In SaaS the provider manages the application, platform and infrastructure, so the customer relies heavily on the provider's security, but the customer is still responsible for its own data, users and access.

Key security concerns and controls

1. Data security: encryption of data in transit (TLS) and at rest, plus data segregation between tenants.
2. Identity and access management: strong authentication (multi-factor), single sign-on and role-based access.
3. Application security: protection against SQL injection, XSS and other web attacks; regular patching and secure development.
4. Availability and backup: redundancy, disaster recovery and SLAs.
5. Compliance and privacy: following regulations such as GDPR; audit logs.
6. Network security: firewalls, intrusion detection and DDoS protection.
7. Vendor lock-in and data location: check where the data is stored and how it can be exported.

Example: with Google Workspace (Gmail, Docs), Google secures the servers and application and encrypts data. The customer enforces two-factor authentication, sets sharing permissions and manages user accounts.

a) Hybrid Clouds

A hybrid cloud combines two or more distinct cloud infrastructures (usually private and public) that stay separate but are bound together by technology allowing data and application portability.

• Sensitive workloads and data stay in the private cloud, while the public cloud supplies extra capacity (cloud bursting) during peaks.
• Benefits: flexibility, cost efficiency, scalability, and security for critical data.
• Challenges: complex integration and management, networking and security consistency.
• Example: a bank keeps customer records in a private cloud but runs its public website and analytics on a public cloud.

c) Data Security

Data security in the cloud is protecting data from unauthorised access, loss and corruption throughout its life cycle.

• Confidentiality: encryption at rest and in transit, key management and access control.
• Integrity: hashing, digital signatures and checksums to detect tampering.
• Availability: backups, replication and disaster recovery.
• Other measures: authentication, auditing and logging, data masking, and compliance with privacy law.
• Risks include data breaches, data leakage, insecure interfaces, insider threats and data remaining after deletion.

Full virtualization

The hypervisor fully emulates the underlying hardware, so an unmodified guest OS runs in the VM, unaware that it is virtualised. Privileged instructions are trapped and translated (binary translation) or handled with hardware assistance (Intel VT-x, AMD-V).

Para-virtualization

The guest OS is modified to know it is virtualised. It uses hypercalls to talk directly to the hypervisor instead of executing privileged instructions, reducing overhead.

Example VMMs

• Full virtualization: VMware ESXi / VMware Workstation, KVM, Microsoft Hyper-V, VirtualBox.
• Para-virtualization: Xen (with modified Linux guests).

Data encryption in cloud computing

Encryption transforms readable plaintext into unreadable ciphertext with a key, so that stolen or intercepted data is useless to attackers. Data must be protected at rest (in storage), in transit (over the network) and, increasingly, in use.

Techniques

1. Symmetric-key encryption: the same key encrypts and decrypts. It is fast and suited to bulk data. Examples: AES (128/256-bit), DES, 3DES. Example: AWS S3 server-side encryption using AES-256.
2. Asymmetric (public-key) encryption: a public key encrypts and a private key decrypts. It is slower but solves key distribution. Examples: RSA, ECC. Example: a client encrypts a session key with the cloud server's public key.
3. Hybrid encryption: asymmetric encryption protects a symmetric key, which in turn encrypts the data. This is how TLS/SSL secures web traffic (HTTPS).
4. Hashing and digital signatures: SHA-256 hashes and signatures provide integrity and authentication (they detect modification of stored data).
5. Homomorphic encryption: computations run on encrypted data without decrypting it, so the cloud never sees the plaintext. It is secure but computationally heavy.
6. Format-preserving / tokenisation / masking: replaces sensitive values (for example card numbers) with tokens or masked values.
7. Client-side and server-side encryption with key management: data is encrypted before upload (client-side) or by the provider (server-side). Keys are stored in a key management service (KMS) or hardware security module (HSM), with rotation.

Best practices: use strong algorithms, protect and rotate keys, keep keys separate from data, use TLS for all transfers, and encrypt backups.

Cloud services are offered in three main models: IaaS, PaaS and SaaS.

As we move from IaaS to SaaS, the provider takes more responsibility, so the user gets less control but more convenience.

SaaS (Software as a Service) delivers software hosted by the provider and accessed through a web browser on a subscription basis.

SaaS lets users start quickly, with no maintenance, from any device. Its drawbacks are limited customisation, dependence on the Internet and the provider, and data-control concerns.

A community cloud is infrastructure shared by several organisations that have common concerns, such as mission, security needs, policy or compliance. It may be managed by the organisations or a third party and hosted on or off premises.

Main benefits

1. Cost sharing: costs are split among members, so it is cheaper than a private cloud.
2. Better security and compliance: it can be built to meet the sector's shared regulations, and is more secure than a public cloud.
3. Collaboration: members can share data, applications and resources easily.
4. Scalability: more capacity than a single private cloud.
5. Customisation: tuned to the community's needs.

Suited organisations and industries

• Government departments and agencies (for example shared e-governance clouds).
• Universities and research institutions sharing computing and data.
• Healthcare providers sharing patient-record systems under privacy rules.
• Banks and financial institutions with common regulations.
• Defence, energy and other organisations with the same security requirements.

A cloud adoption roadmap is a phased plan that guides an organisation in moving its applications, data and operations to the cloud with controlled risk and cost.

Typical phases

1. Assessment and strategy: define business goals, assess current IT, applications and readiness, and build the business case (cost, risk, benefits).
2. Choose models and providers: select service models (IaaS, PaaS, SaaS) and deployment models (public, private, hybrid); evaluate vendors on security, cost, SLA and lock-in.
3. Planning and design: classify workloads, decide the migration approach (rehost, replatform, refactor, replace), and plan security, governance, compliance and network.
4. Pilot / proof of concept: migrate a small, low-risk application to test performance, cost and security.
5. Migration: move applications and data in stages, with testing and training of staff.
6. Optimisation and management: monitor performance and cost, automate, scale and apply security controls.
7. Governance and continuous improvement: manage SLAs, compliance and change, and review the roadmap regularly.

Service-Oriented Architecture (SOA) builds applications from loosely coupled, reusable services that communicate through standard interfaces.

Main roles

1. Service provider: creates and hosts the service and publishes its description (contract/interface, for example WSDL) to the registry.
2. Service registry (broker): a directory (for example UDDI) where providers publish and consumers discover services.
3. Service consumer (requester): finds a service in the registry and binds to it to use its functionality.

Interaction (publish - find - bind)

1. The provider publishes the service description to the registry.
2. The consumer finds a suitable service by querying the registry.
3. The consumer binds to the provider and invokes the service with messages such as SOAP or REST over HTTP.
4. The provider returns the result.

Value to the business: reusability of services, faster development and integration, flexibility to change processes, platform independence, lower cost and better alignment of IT with business needs.

Aneka is a .NET-based Platform-as-a-Service (PaaS) framework from Manjrasoft for building and deploying applications on distributed infrastructure (desktop grids, clusters, data centres and public clouds). It provides a runtime, a set of APIs and tools so that developers can write applications once and run them on many resources.

Design for managing distributed computing

• A pool of nodes runs the Aneka container; the containers together form the Aneka cloud.
• One or more master nodes schedule and coordinate work. Worker nodes execute tasks.
• Applications are expressed through programming models (Task, Thread, MapReduce) and submitted to the cloud. The platform schedules, executes and monitors them.
• Dynamic provisioning adds resources from public clouds (for example Amazon EC2, Azure) when demand rises, and releases them when demand falls.
• Billing, accounting, reservation and quality-of-service (QoS) support are built in.

Core architecture (layered)

1. Fabric layer: the infrastructure (physical or virtual machines) and the Platform Abstraction Layer (PAL) that hides OS differences and gives hardware and node information.
2. Foundation services: membership and configuration, resource reservation, storage management, accounting and billing, reporting and monitoring, licensing.
3. Application services: scheduling, execution, and data-transfer services for each programming model.
4. Programming models and SDK: Task, Thread and MapReduce models, plus management tools and APIs for building and submitting applications.

The container is the basic building block. It hosts services, manages their lifecycle and exposes them to the rest of the cloud, and so makes the design extensible and scalable.

(a) Public and Private Cloud

• Public cloud: resources (servers, storage, applications) are owned by a provider and offered over the internet to many customers on a shared, pay-as-you-go basis, for example AWS, Azure and Google Cloud. Advantages: no upfront capital cost, near-unlimited scalability and no maintenance by the user. Disadvantages: less control, shared infrastructure, and security and compliance concerns.
• Private cloud: infrastructure dedicated to a single organization, hosted on-premises or by a third party. Advantages: greater control, security, customisation and regulatory compliance. Disadvantages: higher cost, in-house expertise needed and limited elasticity.

(b) IaaS (Infrastructure as a Service)

IaaS provides virtualised computing resources over the internet: virtual machines, storage, networks and load balancers. The user manages the operating system, middleware, runtime and applications, while the provider manages the physical hardware, virtualisation and data centre. Billing is by usage. Examples: Amazon EC2, Google Compute Engine, Azure Virtual Machines. Benefits are flexibility, rapid scaling and no hardware purchase. It suits hosting websites, test and development environments, storage, backup and big-data analysis.

(c) Data Security

Data security in the cloud is the set of controls that protect data from unauthorised access, loss, corruption and leakage. Key measures: encryption of data at rest and in transit, strong identity and access management with multi-factor authentication, key management, regular backups and disaster recovery, data masking and isolation between tenants, monitoring and audit logs, and compliance with laws and standards.

(Any two are required; all three are given.)

Cloud security is the set of policies, technologies, controls and procedures that protect cloud data, applications and infrastructure from threats. Security is a shared responsibility: the provider secures the underlying infrastructure, while the customer secures its data, identities, configuration and applications.

Measures to secure cloud applications and services

1. Identity and access management: least privilege, role-based access control, multi-factor authentication and single sign-on.
2. Encryption: TLS for data in transit; AES-based encryption for data at rest; proper key management (KMS/HSM).
3. Network security: firewalls, security groups, virtual private cloud (VPC) segmentation, VPN and intrusion detection/prevention.
4. Secure application development: input validation, secure APIs, code review, vulnerability scanning and timely patching.
5. Monitoring and logging: centralised logs, SIEM, alerts and regular audits.
6. Backup and disaster recovery: regular backups and tested recovery plans.
7. Compliance and governance: adherence to standards such as ISO 27001, GDPR and SLAs that state security terms.

Key threats and mitigation

Together these layered controls (defence in depth), combined with continuous monitoring and staff training, give a secure cloud environment.

Concept

MapReduce is a programming model and framework (introduced by Google, implemented in Hadoop) for processing very large data sets in parallel on a cluster of commodity machines. The user writes two functions, map and reduce; the framework handles partitioning, scheduling, fault tolerance and communication.

Function in distributed computing

It divides a huge job into many small independent tasks that run on different nodes close to where the data is stored (data locality), then combines the results. It scales almost linearly with more nodes, hides parallel-programming complexity, and recovers from failures by re-executing failed tasks.

Procedure of a MapReduce job

1. Input splitting: input data (stored in a distributed file system such as HDFS) is divided into splits.
2. Map: each mapper reads a split and emits intermediate key-value pairs: $map(k_1, v_1) \rightarrow list(k_2, v_2)$.
3. Shuffle and sort: intermediate pairs are partitioned, transferred across the network and grouped by key, giving $(k_2, list(v_2))$. An optional combiner does local aggregation to reduce traffic.
4. Reduce: each reducer processes one key group and emits the result: $reduce(k_2, list(v_2)) \rightarrow list(v_3)$.
5. Output: results are written to the distributed file system.

Example: word count

For the input lines "cloud is big" and "big data is big":

• Map emits: (cloud,1), (is,1), (big,1), (big,1), (data,1), (is,1), (big,1)
• After shuffle: big -> [1,1,1], cloud -> [1], data -> [1], is -> [1,1]
• Reduce sums: (big,3), (cloud,1), (data,1), (is,2)

def map(line):
    for w in line.split():
        emit(w, 1)

def reduce(word, counts):
    emit(word, sum(counts))

Other uses include log analysis, inverted indexes for search, and sorting.

Virtualization creates virtual versions of hardware, so that several isolated environments share one physical machine. Its two major elements are the hypervisor and the virtual machine.

Hypervisor (Virtual Machine Monitor)

Software (or firmware) that creates, runs and manages virtual machines and allocates CPU, memory, storage and I/O among them.

• Type 1 (bare-metal): runs directly on hardware, for example VMware ESXi, Xen, Microsoft Hyper-V.
• Type 2 (hosted): runs on a host operating system, for example VirtualBox, VMware Workstation.

Benefits: efficient resource sharing, strong isolation, central management, live migration, snapshots. Drawbacks: it is a single point of failure and an attack target; Type 2 has extra overhead from the host OS; licensing and management complexity.

Virtual Machine (VM)

A software emulation of a complete computer with its own virtual CPU, memory, disk, network and a guest OS, running on the hypervisor.

Benefits: isolation between workloads, easy backup, cloning and portability, running different operating systems on one host, quick provisioning. Drawbacks: each VM carries a full guest OS (heavy memory and disk use), slower than native hardware, possible performance interference between VMs ("noisy neighbour"), and VM sprawl.

Comparison

In short, the hypervisor provides the virtualization, and VMs are the isolated machines it delivers.

Hypervisor

A hypervisor, or virtual machine monitor (VMM), is software that creates and runs virtual machines by sharing the physical CPU, memory, storage and network among them. It keeps each VM isolated from the others.

Type 2 hypervisor (hosted)

It is installed as an application on a host operating system (Windows, Linux, macOS) and uses the host OS to reach the hardware. Examples: Oracle VirtualBox, VMware Workstation, Parallels.

Advantages

• Easy to install and use, like any application.
• Runs on ordinary desktops and laptops, with no special hardware or dedicated server.
• Wide hardware support through the host OS drivers.
• Good for development, testing, learning and running a second OS.
• Low cost; the host OS can still be used for normal work.

Disadvantages

• Higher overhead and lower performance, because requests pass through the host OS.
• Less secure: a flaw or attack on the host OS affects all VMs.
• If the host OS crashes, all VMs stop.
• Poor scalability and resource management, so unsuitable for large production data centres.
• Limited advanced features such as high availability compared with Type 1.

Cloud computing delivers computing resources over the internet on demand, and its services are grouped into three main service models.

1. Infrastructure as a Service (IaaS): virtual machines, storage, networks and load balancers. The user controls the OS and applications; the provider runs the hardware. Examples: Amazon EC2, Google Compute Engine, Azure VMs.
2. Platform as a Service (PaaS): a ready platform with runtime, databases, middleware and development tools so developers can build and deploy applications without managing servers. Examples: Google App Engine, Azure App Service, Heroku, Aneka.
3. Software as a Service (SaaS): complete applications delivered through a browser and managed by the provider. Examples: Gmail, Google Docs, Salesforce, Microsoft 365.

Other services built on these include Storage as a Service (Dropbox, S3), Database as a Service, Security as a Service, Function as a Service (serverless, AWS Lambda), and Anything/Everything as a Service (XaaS).

Common features are on-demand self-service, broad network access, resource pooling, rapid elasticity and measured (pay-per-use) billing.

Definition

MapReduce is a programming model for processing and generating large data sets in parallel across a cluster. A job has two phases: Map, which turns input records into intermediate key-value pairs, and Reduce, which merges all values that share a key into the final result. The framework (for example Hadoop) handles splitting, scheduling, shuffling, and recovery from node failure.

Uses

• Large-scale data processing such as counting words and aggregating statistics.
• Log analysis and web-server click-stream analysis.
• Building inverted indexes for search engines.
• Distributed sorting and searching (grep).
• Data mining, machine learning and recommendation processing on big data.
• Graph processing such as PageRank, and ETL on data warehouses.
• Processing scientific and sensor data.

Its strengths are scalability on cheap hardware, fault tolerance, and a simple model that hides the complexity of parallel programming.

Web services

A web service is a software component that offers functionality over a network using standard web protocols, so that different applications can interoperate regardless of platform or language. Typical standards are HTTP, XML/JSON, SOAP, REST, WSDL and UDDI. Cloud providers expose their resources (storage, compute) as web services/APIs, for example Amazon S3 and EC2.

On-demand resource provisioning

Users acquire computing resources (servers, storage, bandwidth) automatically, whenever needed, through a self-service portal or API, without human interaction with the provider. Resources are ready in minutes and released when finished, so there is no wait to buy and install hardware.

Scalability

The ability to grow or shrink resources as workload changes. Vertical scaling adds power to one machine; horizontal scaling adds more machines. In the cloud this can be automatic (auto-scaling), which keeps performance steady during peaks and avoids waste at low load.

Pay-as-you-go pricing

Customers pay only for the resources actually consumed (per hour, per GB or per request), like electricity. This removes upfront capital expense, turns it into operational expense, and lowers cost for variable or unpredictable workloads.

Bucket in Amazon S3

Amazon Simple Storage Service (S3) is an object storage service. A bucket is the top-level container in which objects (files plus metadata) are stored. Every object belongs to exactly one bucket and is identified by a key (its name) within it. Bucket names are globally unique across all AWS accounts and regions (3 to 63 characters, lowercase letters, numbers and hyphens), and a bucket is created in a chosen region. Access control, versioning, lifecycle and logging are configured per bucket.

Addressing of a bucket

A bucket and its objects are reached over HTTP(S) in two URL styles:

1. Virtual-hosted style: the bucket name is part of the host name.

$\texttt{https://<bucket-name>.s3.<region>.amazonaws.com/<object-key>}$

Example: https://mybucket.s3.ap-south-1.amazonaws.com/photos/pic.jpg

2. Path style: the bucket name is in the path.

$\texttt{https://s3.<region>.amazonaws.com/<bucket-name>/<object-key>}$

Example: https://s3.ap-south-1.amazonaws.com/mybucket/photos/pic.jpg

An S3 URI such as s3://mybucket/photos/pic.jpg is used by tools and SDKs. Because the name is globally unique and forms part of a DNS name, it must follow DNS naming rules.

(a) Parallel Computing

Parallel computing performs many calculations at the same time by splitting a problem into smaller parts that run simultaneously on multiple processors or cores. Its aim is to solve large problems faster than serial execution. Forms include bit-level, instruction-level, data and task parallelism. Architectures include shared-memory multiprocessors and distributed-memory clusters, programmed with models such as threads, OpenMP, MPI and MapReduce. Speed-up is limited by the serial fraction of the program (Amdahl's law) and by communication and synchronisation overhead. It underlies cloud and high-performance computing.

(b) PaaS (Platform as a Service)

PaaS provides a complete development and deployment platform in the cloud: operating system, runtime, web server, database, middleware and tools. Developers write and deploy code while the provider manages servers, storage, networking, scaling and patching. Benefits: faster development, lower cost, automatic scaling and focus on the application. Drawbacks: vendor lock-in and less control. Examples: Google App Engine, Azure App Service, Heroku, AWS Elastic Beanstalk.

(c) Microsoft Azure

Azure is Microsoft's public cloud platform offering IaaS, PaaS and SaaS services through global data centres. Main services: Virtual Machines, App Service, Azure Storage (Blob, Queue, Table, File), SQL Database and Cosmos DB, Azure Functions, networking, Active Directory identity, and AI and analytics tools. It integrates well with Windows Server, .NET and Microsoft tools, supports hybrid cloud and offers pay-as-you-go pricing.

(Any two are required; all three are given.)

Definition

Cloud computing is the on-demand delivery of computing resources (servers, storage, databases, networking, software) over the internet with pay-as-you-go pricing. NIST defines it as a model for convenient, on-demand network access to a shared pool of configurable resources that can be rapidly provisioned with minimal management effort. Its five characteristics are on-demand self-service, broad network access, resource pooling, rapid elasticity and measured service.

Challenges

1. Security and privacy: sensitive data is held by a third party.
2. Performance and latency: dependence on network speed and shared resources.
3. Downtime and availability: outages affect all customers; internet connectivity is required.
4. Vendor lock-in and interoperability: moving between providers is difficult because of proprietary APIs.
5. Compliance and legal issues: data location and laws such as GDPR.
6. Cost management: unexpected bills from over-provisioning and data transfer.
7. Limited control and visibility over the infrastructure.
8. Migration complexity and skill shortage.
9. Energy consumption of data centres.

Security issues

• Data breaches and data loss.
• Insecure interfaces and APIs.
• Account or service hijacking through stolen credentials.
• Insider threats from provider or customer staff.
• Denial of Service (DDoS) attacks.
• Multi-tenancy risks: attacks across shared hardware, VM escape, side channels.
• Misconfiguration such as publicly open storage.
• Malware and shared-technology vulnerabilities.
• Lack of transparency and control, and weak data deletion.

Remedies: encryption, strong identity management with MFA, network segmentation, regular audits, backups, clear SLAs, and compliance with standards. Security is a shared responsibility between provider and customer.

Definition

Virtualization is the technology that creates a virtual (logical) version of computing resources, such as servers, storage, networks or operating systems, abstracted from the physical hardware. It lets one physical resource be divided into several isolated virtual ones, or several resources be combined into one, and it is the foundation of cloud computing.

Types of virtualization

1. Hardware (server) virtualization A hypervisor divides one physical server into many virtual machines, each with its own OS. Advantages: high hardware utilisation, lower cost, isolation, easy migration and backup. Disadvantages: performance overhead, hypervisor is a single point of failure, VM sprawl.

2. Operating system (container) virtualization The OS kernel is shared and the host runs isolated containers (Docker, LXC). Advantages: lightweight, fast start-up, high density. Disadvantages: all containers must use the same kernel; weaker isolation than VMs.

3. Storage virtualization Physical storage devices are pooled and presented as one logical storage (SAN, RAID, software-defined storage). Advantages: easier management, better utilisation, simple backup and migration. Disadvantages: complex set-up, performance overhead from the mapping layer.

4. Network virtualization Physical networks are combined or split into virtual networks (VLAN, VPN, SDN). Advantages: flexible, isolated networks, central control, lower hardware cost. Disadvantages: configuration complexity, security and troubleshooting difficulty.

5. Desktop virtualization Desktops run on central servers and are accessed remotely (VDI). Advantages: centralised management and security, access from any device. Disadvantages: needs a reliable network, high server cost.

6. Application virtualization An application runs in an isolated environment without installation on the local OS (Citrix, App-V). Advantages: no conflicts, easy deployment and updates. Disadvantages: not all apps are supported; performance and licensing issues.

7. Data virtualization Data from different sources is accessed through one logical layer without moving it. Advantages: unified real-time view, reduced duplication. Disadvantages: query performance depends on the sources.

Before adopting cloud computing, an organization should weigh the following factors.

1. Business needs and cost: compare capital expense with pay-as-you-go cost, including long-term total cost of ownership, data transfer fees and hidden charges.
2. Security and privacy: encryption, access control, and the provider's security certifications.
3. Compliance and legal requirements: data residency, laws such as GDPR, and industry regulations.
4. Reliability and availability: the provider's uptime record, SLA guarantees, backup and disaster recovery.
5. Performance and network: latency, bandwidth and dependable internet connectivity.
6. Scalability and flexibility: ability to grow or shrink with demand.
7. Vendor lock-in and interoperability: portability of data and applications, open standards, exit strategy.
8. Integration and migration: compatibility with existing systems and effort needed to move.
9. Choice of model: public, private, hybrid or community cloud, and IaaS, PaaS or SaaS.
10. Skills and management: staff training, change management and support from the provider.

A pilot project and a clear risk assessment help the decision.

Service-Oriented Architecture (SOA)

SOA is a design style in which software is built from loosely coupled, reusable, self-contained services that communicate through standard, well-defined interfaces (SOAP/WSDL or REST over HTTP). A service-oriented design fits cloud computing because the cloud itself exposes resources and applications as services (IaaS, PaaS, SaaS).

Design and implementation using SOA

1. Identify and define services: break business functions into independent services, each with a clear contract (interface).
2. Service roles: the provider implements and hosts the service; the registry lists service descriptions; the consumer finds and calls it (publish, find, bind).
3. Communication: services exchange messages using standard protocols and formats (HTTP, XML, JSON), often through an enterprise service bus.
4. Deployment in the cloud: services are deployed as virtual machines, containers or functions; the cloud supplies elastic scaling, load balancing and monitoring.
5. Composition and orchestration: several services are combined into larger business processes (for example with BPEL).
6. Governance and security: authentication, authorisation, SLAs and monitoring for each service.

Benefits of SOA in cloud computing

• Reusability of services and faster development.
• Platform and language independence (interoperability).
• Loose coupling, so services can change independently.
• Scalability, since individual services can be scaled on demand.
• Easier integration of legacy and new systems.
• Lower cost and better maintainability.
• Flexibility and agility to adapt to changing business needs.

Benefits of virtualization in cloud computing

• Server consolidation: many VMs run on one physical server.
• Isolation: each VM is separate, so a failure or attack in one does not affect others.
• Rapid provisioning and elasticity: new VMs are created in minutes from templates.
• Multi-tenancy: many customers share the same hardware securely.
• Flexibility and portability: VMs can be moved (live migration), cloned, backed up with snapshots.
• Disaster recovery and high availability: VMs are easily restarted on other hosts.
• Testing and development in safe, disposable environments.

Improving resource utilisation

A dedicated physical server often uses only 10 to 15 percent of its capacity. Virtualization places several workloads on one host and dynamically allocates CPU, memory and storage as needed, raising utilisation to 60 to 80 percent. Idle capacity is shared, and resources are reallocated where demand is highest.

Reducing operational costs

• Fewer physical servers mean lower hardware purchase cost.
• Lower electricity, cooling and data-centre space costs.
• Centralised management reduces administration effort and staff cost.
• Faster deployment and recovery reduce downtime losses.
• Pay-per-use sharing means customers pay only for what they use.

How MapReduce works

MapReduce processes a large data set in parallel on a cluster in these steps.

1. Split: the input is divided into blocks (splits) stored across nodes (for example HDFS).
2. Map: each mapper processes one split and emits intermediate key-value pairs, $(k_1, v_1) \rightarrow list(k_2, v_2)$.
3. Shuffle and sort: pairs are grouped by key and sent to reducers.
4. Reduce: each reducer aggregates the values of a key, $(k_2, list(v_2)) \rightarrow list(v_3)$.
5. Output: results are written back to the file system.

A master node schedules tasks, moves computation to the data, and re-runs tasks that fail. For example, in word count the map emits (word, 1) and the reducer sums the counts.

Advantages over traditional methods

• Scalability: processes terabytes to petabytes by adding cheap commodity nodes; a traditional single-server or RDBMS approach hits hardware limits.
• Parallelism: many tasks run at once, so jobs finish much faster.
• Data locality: code moves to the data, which reduces network traffic.
• Fault tolerance: failed tasks are re-executed automatically and data is replicated.
• Simplicity: the programmer writes only map and reduce; parallelisation and distribution are automatic.
• Cost effective: uses inexpensive hardware.
• Handles unstructured data such as logs and text.

Key strategies for data security in the cloud:

1. Encryption: encrypt data at rest (for example AES-256) and in transit (TLS/SSL). Keep the keys under the customer's control using a key management service or hardware security module.
2. Identity and access management (IAM): least privilege, role-based access, multi-factor authentication, and prompt removal of unused accounts.
3. Data classification and masking: identify sensitive data, apply masking, tokenisation or anonymisation.
4. Backup and disaster recovery: regular, encrypted, geographically separate backups and tested recovery plans.
5. Network security: firewalls, VPN, private networks (VPC), network segmentation and intrusion detection.
6. Monitoring and auditing: logging, SIEM, alerts on unusual access, and regular audits.
7. Data loss prevention (DLP) tools to stop leakage.
8. Secure configuration: avoid public storage buckets, patch systems and scan for vulnerabilities.
9. Compliance and SLAs: choose providers certified to ISO 27001, SOC 2 and similar standards; define data location, ownership and deletion in contracts.
10. Secure deletion and user training to reduce human error.

Role of cloud computing

Business applications: customer relationship management (Salesforce), enterprise resource planning, accounting and payroll, e-mail and collaboration (Microsoft 365, Google Workspace), e-commerce hosting, data analytics, backup and disaster recovery, and software development and testing.

Consumer applications: web mail (Gmail), photo and file storage (Google Drive, iCloud, Dropbox), social networks, video and music streaming (YouTube, Netflix), online office tools, and online gaming.

Benefits for businesses

Scalability

• Resources grow or shrink on demand, automatically if configured (elasticity).
• A business can handle seasonal peaks, such as festival sales, without buying extra servers.
• New services and branches can be launched quickly.

Cost-effectiveness

• No large upfront hardware investment; costs become operating expenses.
• Pay-as-you-go billing means no payment for idle capacity.
• Lower costs for maintenance, power, space and IT staff.
• Updates and maintenance are handled by the provider.

Additional benefits are anywhere access, easy collaboration, built-in backup and faster time to market.

Cloud support for scientific applications

Scientific applications such as climate modelling, genomics, astronomy, physics simulation and drug discovery need large computing and storage capacity. Cloud platforms support them by:

• Providing on-demand access to thousands of virtual machines, GPUs and large storage, with no need to own a cluster.
• Offering parallel and distributed frameworks (MapReduce/Hadoop, Spark, MPI clusters) and workflow engines.
• Giving scalable data storage and transfer for huge data sets.
• Enabling collaboration and sharing of data and results among research groups worldwide.
• Supplying ready-made software environments through VM and container images, which makes experiments reproducible.
• Supporting examples such as Amazon EC2 clusters, Azure and Aneka for scientific workloads.

Benefits of cloud for HPC in research

• Cost savings: pay per use instead of building and maintaining a supercomputer.
• Elastic scalability: thousands of cores for a short time, then release them.
• Fast access: start in minutes instead of waiting in queues or for procurement.
• Access for small institutions that cannot afford a cluster.
• Latest hardware (GPUs, accelerators) without upgrade cost.
• Reduced maintenance and administration.
• Reliability through replication and backup.

Limitations are network latency, data-transfer cost and overhead from virtualization for tightly coupled jobs.

Thread

A thread is the smallest unit of execution within a process. Threads of one process share the same memory (code, data, heap) and resources, but each has its own program counter, registers and stack. They are lighter than processes and allow concurrent execution inside one application.

Thread programming in the cloud

Multithreaded programming lets an application split work into concurrent tasks to use multiple cores and many nodes. In the cloud:

• On a single VM, threads use the cores with standard libraries (Java threads, POSIX threads, .NET Thread).
• In distributed cloud platforms such as Aneka, the Thread programming model offers a distributed thread abstraction that looks like a local thread but is executed on a remote node. The developer creates thread objects, the runtime sends them to the cloud, runs them, and collects results.
• Typical steps: define the work in a thread class; create and configure the application; start the threads (Start); wait for completion (Join); read the results.
• Threads in the cloud need synchronisation (locks, semaphores) for shared data, and care to avoid deadlock and race conditions. Because remote threads do not share memory, data is passed through serialisation and messages.
• Thread-based parallelism suits problems that are divisible into independent units (parameter sweeps, simulations).

How virtualization is transforming the IT industry

• Server consolidation: fewer physical servers, lower hardware, power and space costs.
• Agility: servers are provisioned in minutes, speeding development and deployment.
• Efficient use of resources through sharing and dynamic allocation.
• Business continuity: live migration, snapshots and quick disaster recovery.
• New models: it enables cloud services, DevOps, containers and virtual desktops.
• Green IT by reducing energy consumption.

Significance in cloud computing

Virtualization is the enabling technology of the cloud. It provides resource pooling and multi-tenancy (many users on shared hardware with isolation), elasticity (resources added or removed on demand), hardware independence and portability of VMs, metering for pay-per-use billing, and automation of provisioning. Without it, IaaS and the economics of the cloud would not be possible.

Critical aspects of application security in the cloud

1. Authentication and authorisation: verify users and services, and give only the needed rights.
2. Data protection: encryption in transit and at rest, and secure handling of secrets and keys.
3. Secure software development life cycle: security considered in design, coding, testing and deployment (DevSecOps).
4. API and interface security: APIs are the main entry points and must be authenticated and validated.
5. Vulnerability and patch management: the application, libraries, OS and containers must be kept updated.
6. Isolation and multi-tenancy: protection from other tenants on shared infrastructure.
7. Logging, monitoring and incident response.
8. Compliance and configuration management.

Techniques used to secure hosted applications

• Input validation and output encoding to prevent SQL injection, cross-site scripting (XSS) and command injection.
• Strong authentication: multi-factor authentication, OAuth 2.0, OpenID Connect, single sign-on, session management and token expiry.
• Role-based access control and least privilege.
• Encryption: TLS for all traffic; encrypted databases and storage.
• Web application firewall (WAF) and DDoS protection.
• Secure API gateway with rate limiting and authentication.
• Code review, static and dynamic testing (SAST/DAST), penetration testing and dependency scanning.
• Container and VM hardening, image scanning, and network segmentation.
• Secrets management instead of hard-coded credentials.
• Continuous monitoring, logging and SIEM, with alerts and a response plan.
• Regular patching, backups and security audits.

These measures, applied in layers, reduce the chance that a vulnerability is exploited.

Community cloud

A community cloud is a cloud infrastructure shared by several organizations with common concerns, for example the same mission, security needs, policy or compliance requirements. It may be owned and managed by one or more of the member organizations, by a third party, or by a combination, and may be on or off premises. Examples are a cloud shared by government departments, by banks, by hospitals, or by universities and research institutions.

Difference from public and private clouds

Situations where it is beneficial

• Organizations with common regulatory or compliance requirements, such as banks, healthcare and government.
• When they want shared costs of infrastructure, security and maintenance instead of each building its own private cloud.
• For collaboration and data sharing among partners, for example joint research projects between universities.
• When a public cloud is not trusted enough but a private cloud is too costly.
• When members need common applications or platforms (for example e-governance services) with consistent policies.

Its limitations include shared bandwidth and capacity, the need for governance agreements between members, and less scalability than the public cloud.