Introduction to Cloud Computing

Load balancing is the process of distributing incoming network traffic or workload evenly across several servers or resources. It ensures that no single server is overloaded, which improves response time, throughput, availability and fault tolerance. In the cloud it supports elasticity, because new instances are added to the pool as demand grows.

How load balancing is done

• A load balancer (hardware or software, for example AWS Elastic Load Balancer, HAProxy or Nginx) sits in front of a pool of servers. Clients send requests to its address, and it forwards each request to a chosen server.
• Health checks detect failed servers, and traffic is then sent only to healthy ones.
• Static algorithms: Round robin, weighted round robin and IP hash.
• Dynamic algorithms: Least connections, least response time and resource-based (CPU or memory) selection.
• Session persistence (sticky sessions) can keep a user on the same server.
• DNS-based and global server load balancing distribute traffic across data centres and regions.
• Auto-scaling works with the load balancer to add or remove instances.
• Load can also be balanced at layer 4 (TCP) or layer 7 (HTTP content-based).

Cloud computing is a model for delivering computing resources (servers, storage, databases, networking, software and analytics) over the Internet on demand. Users get them with minimal management effort and pay only for what they use (pay-as-you-go). NIST describes five essential characteristics: on-demand self-service, broad network access, resource pooling, rapid elasticity and measured service.

Cloud deployment models

(a) Public cloud

• Owned and operated by a third-party provider and shared by many customers over the Internet.
• Examples: AWS, Microsoft Azure, Google Cloud.
• Advantages: no capital cost, high scalability, no maintenance.
• Disadvantages: less control and customisation, and security and compliance concerns.

(b) Private cloud

• Used by a single organisation. It may be hosted on-premises or by a third party.
• Examples: VMware or OpenStack based internal clouds in banks and government.
• Advantages: greater security, control and compliance.
• Disadvantages: high cost and the organisation does the management.

(c) Community cloud

• Shared by several organisations with common concerns (security, compliance, mission), for example government agencies, universities or hospitals.
• Costs are shared. It may be managed internally or by a third party.

(d) Hybrid cloud

• A combination of two or more clouds (typically private and public) joined by technology that allows data and applications to move between them.
• Sensitive workloads stay private, and demand peaks are handled by "cloud bursting" to the public cloud.
• It is more complex to manage.

Parallel efficiency measures how well the processors in a parallel system are used. If $T_1$ is the time on one processor and $T_p$ the time on $p$ processors:

$\text{Speedup } S = \frac{T_1}{T_p}, \qquad \text{Efficiency } E = \frac{S}{p} = \frac{T_1}{p\,T_p}$

$E = 1$ (100%) is ideal. In practice $E<1$ because of communication, synchronisation, load imbalance and serial portions (Amdahl's law). In MapReduce, parallel efficiency means that adding more nodes reduces the job time almost in proportion. Map tasks run independently on data blocks stored locally, and the framework handles scheduling, data locality, failure recovery and load balancing. So efficiency remains high for large data sets.

Comparison of thread, task and MapReduce

Virtualization is the creation of a virtual (rather than physical) version of a resource, such as a server, storage device, network or operating system. A virtualization layer (the hypervisor) lets many virtual machines (VMs) share one physical machine, each isolated from the others.

Why virtualization suits cloud computing

• Resource pooling and multi-tenancy: One physical server hosts many customers' VMs.
• Elasticity: VMs can be created, resized or removed in minutes.
• Better utilisation and lower cost: Server consolidation reduces hardware, power and space.
• Isolation and security: Failures and attacks in one VM do not affect others.
• Live migration and fault tolerance: VMs move between hosts without downtime.
• Easy provisioning, backup and disaster recovery through VM images and snapshots.
• Hardware independence and legacy support.

Types of virtualization

1. Hardware (server) virtualization: A hypervisor creates VMs. A Type 1 (bare-metal) hypervisor such as Xen, ESXi or Hyper-V runs on the hardware. A Type 2 (hosted) hypervisor such as VirtualBox runs on an OS. Variants are full virtualization, para-virtualization and hardware-assisted virtualization.
2. Operating-system-level virtualization (containers): Many isolated user spaces share one kernel, for example Docker and LXC.
3. Storage virtualization: Several physical storage devices appear as one pool (SAN, RAID).
4. Network virtualization: Physical networks are divided into logical ones (VLAN, VPN, SDN).
5. Desktop virtualization (VDI): Desktops run on central servers and are accessed remotely.
6. Application virtualization: An application runs without being installed on the client (for example Citrix, JVM).
7. Memory and data virtualization are also used.

Cloud adoption is the process by which an organisation moves its applications, data and IT operations from traditional on-premises infrastructure to cloud services, whether partially or wholly.

Factors leading organisations to adopt the cloud

• Cost savings: Capital expense becomes operating expense (pay-as-you-go), and there is no hardware maintenance.
• Scalability and elasticity: Resources scale up and down with demand.
• Speed and agility: Resources are provisioned in minutes, so products reach the market sooner.
• Reliability and disaster recovery: Redundancy, backups and high availability.
• Access to advanced technology: AI, analytics and big data tools without owning them.
• Mobility and collaboration: Access from anywhere on any device.
• Focus on core business: The provider handles IT management.
• Security and compliance offered by big providers, and competitive pressure and green IT goals.

Platform-as-a-Service (PaaS) gives developers a ready environment to build, test, deploy and manage applications without managing the underlying servers, storage, networks or OS. Examples: Google App Engine, Microsoft Azure App Service, AWS Elastic Beanstalk, Heroku.

Services provided under PaaS

• Development frameworks and tools: Programming languages, runtimes, SDKs, IDE integration and APIs.
• Application hosting and deployment: Automatic deployment, versioning and scaling.
• Database and storage services: Managed relational and NoSQL databases.
• Middleware services: Messaging queues, caching, authentication and integration.
• Testing and collaboration: Testing environments, team development and version control.
• Business intelligence and analytics services.
• Management services: Monitoring, load balancing, logging and security.

Benefits: Faster development, lower cost, no infrastructure management and automatic scaling. Limitation: vendor lock-in.

Security-as-a-Service (SECaaS) is a model in which a provider delivers security services over the cloud on a subscription basis, so the customer does not install or maintain security infrastructure.

Features and services

• Identity and access management (IAM): Single sign-on, multi-factor authentication and role-based access.
• Anti-virus and anti-malware: Cloud-based scanning and updates.
• Intrusion detection and prevention (IDS/IPS) and firewall as a service.
• Email security: Spam and phishing filtering.
• Web security: Web filtering and secure gateways.
• Data encryption and data loss prevention (DLP).
• Security information and event management (SIEM): Log collection, monitoring and alerts.
• Vulnerability scanning and penetration testing.
• Business continuity and disaster recovery, and compliance and audit reporting.
• Advantages: Lower cost, up-to-date protection, expert staff, scalability and quick deployment.

Designing a cloud security architecture means building security into every layer of the cloud (network, host, application and data) in line with the shared responsibility between provider and customer.

Guidelines and principles

1. Defence in depth: Use multiple layers of controls (firewalls, IDS, encryption, access control).
2. Least privilege and need-to-know: Give users and services only the access they need.
3. Strong identity and access management: Multi-factor authentication, role-based access and federated identity.
4. Data protection: Encrypt data at rest and in transit, manage keys securely, classify data and prevent leakage.
5. Network security: Segmentation, VPN, virtual private clouds and secure APIs.
6. Separation and isolation of tenants (hypervisor and VM isolation).
7. Continuous monitoring, logging and auditing: Detect incidents early and keep audit trails.
8. Compliance and governance: Follow laws and standards (ISO 27001, GDPR, PCI-DSS) and agree SLAs.
9. Secure the software lifecycle: Patch management, vulnerability testing and secure coding.
10. Resilience: Backup, disaster recovery, redundancy and incident response planning.
11. Zero trust and "fail secure": Never trust by default, verify every request, and fail into a safe state.
12. Clear shared-responsibility model and risk assessment.

A bucket in Amazon S3 (Simple Storage Service) is a container that holds objects (files and their metadata). Every object is stored in a bucket and is identified by a key. Bucket names are globally unique, and a bucket is created in a chosen AWS region. It provides durable, scalable object storage, access control and versioning.

Creating a bucket (AWS Management Console)

1. Sign in to the AWS console and open the S3 service.
2. Click Create bucket.
3. Enter a globally unique bucket name (lower case, 3-63 characters) and select an AWS Region.
4. Set Object Ownership and Block Public Access settings (public access is blocked by default).
5. Optionally enable versioning, default encryption and tags.
6. Click Create bucket.
7. Open the bucket and use Upload to add objects. Set permissions with bucket policies or ACLs.

CLI equivalent

aws s3 mb s3://my-unique-bucket --region ap-south-1

Google App Engine (GAE) is a PaaS from Google Cloud that lets developers build and host web applications and APIs on Google's infrastructure without managing servers.

Use for development

• Write the application in a supported language (Python, Java, Go, Node.js, PHP, Ruby) with standard frameworks.
• Test locally with the SDK or Cloud CLI development server.
• Define settings in a configuration file (app.yaml), such as runtime, scaling and URL handlers.
• Use the integrated services: Datastore/Firestore, Cloud SQL, Memcache, Task Queues, Cron, Mail, Users/Identity and Cloud Storage.

Use for hosting

• Deploy with gcloud app deploy. The platform hosts the app at https://<project>.appspot.com or a custom domain.
• Automatic scaling (up and down, even to zero instances), load balancing and traffic splitting between versions.
• Choose the standard environment (sandboxed, quick scale) or the flexible environment (Docker containers).
• Monitoring and logging through Cloud Monitoring, and managed HTTPS and security.
• Pay only for the resources used, with a free quota.

Microsoft Azure is Microsoft's cloud platform providing IaaS, PaaS and SaaS services through global data centres.

Main services

• Compute: Virtual Machines, App Service, Azure Functions (serverless), Container Instances, Kubernetes Service and Cloud Services (web and worker roles).
• Storage: Blob, Table, Queue and File storage, and Disk storage.
• Databases: Azure SQL Database, Cosmos DB (NoSQL) and managed MySQL or PostgreSQL.
• Networking: Virtual Network, Load Balancer, Application Gateway, VPN Gateway, CDN and Traffic Manager.
• Identity and security: Azure Active Directory (Entra ID), Key Vault and Security Center.
• Analytics and big data: HDInsight, Synapse Analytics and Stream Analytics.
• AI and machine learning: Azure Machine Learning and Cognitive Services.
• IoT: IoT Hub.
• DevOps and management: Azure DevOps, Monitor and Resource Manager.
• Integration: Service Bus and Logic Apps.

Remote ECG monitoring means a patient's heart signals are captured continuously by a wearable or bedside sensor and analysed remotely by doctors. Cloud computing suits this scientific and healthcare application well.

How the cloud supports it

1. Data acquisition: ECG sensors send signals through a smartphone or gateway to the cloud over the Internet.
2. Scalable storage: Large streams of ECG data are stored in cloud storage or databases, with backups and long-term records.
3. Processing and analysis: Compute services (or MapReduce and machine learning) filter noise, detect arrhythmias or abnormal beats in near real time, and trigger alerts.
4. Software as a service: Doctors view waveforms and reports in a web portal on any device.
5. Elasticity: Resources grow with the number of patients, and cost is pay-per-use.
6. Alerts and emergency response: The system sends SMS or app alerts to doctors and family when an abnormality is found.
7. Security and privacy: Encryption and access control protect patient data.

Example: Aneka or AWS/Azure IoT hubs can implement this workflow. The result is lower cost, 24x7 monitoring, early detection and wide availability.

Cloud computing is the delivery of computing resources (servers, storage, databases, networking, software, analytics) over the Internet on demand, with pay-per-use pricing, rapid elasticity and little management effort from the user.

Cloud service models

(a) Infrastructure-as-a-Service (IaaS)

• The provider supplies virtualised computing resources: virtual machines, storage and networks.
• The user manages the OS, middleware, runtime, applications and data.
• Examples: Amazon EC2, Azure Virtual Machines, Google Compute Engine.
• Suitable for system administrators and for hosting, test environments and backup.

(b) Platform-as-a-Service (PaaS)

• The provider supplies a platform (runtime, database, middleware, tools) to develop, test and deploy applications.
• The user manages only the application and its data.
• Examples: Google App Engine, Azure App Service, Heroku, AWS Elastic Beanstalk.
• Suitable for developers.

(c) Software-as-a-Service (SaaS)

• Complete applications are delivered through a browser on subscription. The provider manages everything.
• Examples: Gmail, Google Docs, Salesforce, Microsoft 365, Zoom.
• Suitable for end users.

Other models are sometimes added: Security-as-a-Service, Database-as-a-Service and Function-as-a-Service (serverless).

Thread: A thread is the smallest unit of CPU execution within a process. Threads of one process share the same memory, code and resources, but each has its own stack and program counter.

Why multi-threading is used

• Improves responsiveness (the user interface stays active while background work runs).
• Uses multi-core CPUs to run work in parallel and raise throughput.
• Lower overhead than processes, because creation and context switching are cheaper and data is shared easily.
• Allows overlap of I/O waiting with computation.
• Suited to servers handling many requests at once.

Thread programming in the cloud

• A cloud application (for example one on Aneka or a VM on EC2) is divided into independent work units, and each unit runs on a thread. A thread pool in the runtime schedules them across the cores or nodes.
• In Aneka's Thread Programming Model, the developer uses an AnekaThread object, which has an API like System.Threading.Thread. Aneka runs the thread transparently on a remote node. The steps are:
  1. Create the thread with the work method.
  2. Start the thread; the cloud scheduler sends it to an available node.
  3. Wait for it with Join/WaitForCompletion and collect results.
  4. Handle failures through resubmission.
• Because the threads are distributed, the data they need must be passed explicitly (shared memory is not available).
• Synchronisation primitives (locks, semaphores) are used on one machine to avoid race conditions and deadlock.

Typical use: image rendering, Monte Carlo simulation and parameter sweeps, where tasks are independent.

Implementation levels of virtualization

1. Instruction Set Architecture (ISA) level: The ISA of the guest is emulated on a different host ISA. Interpretation or dynamic binary translation is used (for example Bochs, QEMU). It is flexible but slow.
2. Hardware abstraction level: A hypervisor (VMM) virtualises CPU, memory and I/O directly on the hardware, giving each VM its own virtual hardware (Xen, VMware ESXi, KVM). It is the most common level in clouds, with good performance.
3. Operating system level: One OS kernel hosts multiple isolated containers or virtual environments (Docker, LXC, OpenVZ). It is lightweight with minimal overhead, but all share the same kernel.
4. Library (user-API) level: Applications are virtualised by emulating an API or library interface (for example Wine, vCUDA).
5. Application level: A process-level virtual machine such as the JVM or .NET CLR runs an application independent of the OS.

Benefits of virtualization

• Better hardware utilisation and server consolidation.
• Lower hardware, power, cooling and space costs.
• Easy and fast provisioning, scaling and cloning.
• Isolation of workloads, and improved security and fault containment.
• Live migration, high availability and disaster recovery through snapshots.
• Support for several operating systems on one machine, and legacy applications.
• Simple testing and development environments.
• Centralised management and green IT.

Major advantages of cloud computing

• Cost saving: No upfront hardware investment; pay-as-you-go. Maintenance and energy costs fall.
• Scalability and elasticity: Resources scale up or down with demand.
• Accessibility and mobility: Access from anywhere with an Internet connection and any device.
• Reliability and availability: Redundancy, backups and disaster recovery.
• Speed and agility: Quick provisioning shortens time to market.
• Automatic updates and maintenance by the provider.
• Collaboration: Shared documents and real-time teamwork.
• Security: Expert protection and compliance from large providers.
• Focus on core business, and access to advanced technologies such as AI and big-data analytics.
• Environmentally friendly: Shared resources use less energy.

Service-Oriented Architecture (SOA) is a design style in which software functions are packaged as loosely coupled, reusable services with well-defined interfaces, communicating over a network using standard protocols (SOAP, REST, WSDL, XML/JSON). A provider publishes a service, and consumers find (registry) and bind to it.

Role of SOA in developing cloud applications

• Foundation of cloud service models: IaaS, PaaS and SaaS are all delivered as services, so SOA is a natural basis for cloud design.
• Loose coupling: Components can be changed, scaled or replaced independently.
• Reusability and composition: Existing services are combined (mashups, workflows) to build new applications quickly.
• Interoperability: Standard interfaces allow different platforms and languages to work together across clouds.
• Scalability and elasticity: Each service can be scaled on its own.
• Location transparency: Services can be discovered and run anywhere in the cloud.
• Easier integration of legacy and cloud systems, and faster development with lower cost.
• It supports pay-per-use billing of services, and modern microservices follow the same idea.

Challenges and risks in cloud security

• Data breaches and data loss: Sensitive data can be exposed or lost through attack, misconfiguration or provider failure.
• Loss of control and visibility: The customer does not control the infrastructure.
• Multi-tenancy risks: Weak isolation may let one tenant access another's data (side-channel and VM escape attacks).
• Insecure interfaces and APIs, and misconfigured storage.
• Account hijacking and insider threats: Stolen credentials or malicious employees.
• Denial-of-service attacks.
• Vendor lock-in and lack of portability.
• Compliance and legal issues: Data location, privacy laws and jurisdiction.
• Shared-technology vulnerabilities in the hypervisor or software.
• Data remnants after deletion, and weak encryption or key management.
• Service availability: Outages affect the customer.
• Risk of unclear shared-responsibility boundaries.

Data security

• Encryption of data at rest (disk, database) and in transit (SSL/TLS, VPN), with careful key management (KMS or customer-held keys).
• Access control: Identity and access management, role-based access, multi-factor authentication and least privilege.
• Data classification, masking and tokenisation for sensitive data.
• Backups, replication and disaster recovery for availability and integrity.
• Data loss prevention, secure deletion and monitoring of access with audit logs.
• Compliance controls and legal agreements (SLAs) covering location and handling.

Application security

• Secure development lifecycle: Secure coding, code review and vulnerability testing (OWASP top-10 protection).
• Web application firewalls, API gateways and input validation to stop injection and XSS.
• Authentication and authorisation for users and services (OAuth, tokens).
• Patch management and regular penetration tests.
• Isolation by containers, VMs and virtual networks, and monitoring with intrusion detection and logging.
• Responsibilities are divided between provider and customer under the shared responsibility model.

Amazon EC2 (Elastic Compute Cloud) is an AWS IaaS service that provides resizable virtual servers, called instances, in the cloud. Users launch instances in minutes and pay only for the time used.

Features

• Instance types: General purpose, compute-optimised, memory-optimised, storage-optimised and GPU instances.
• Amazon Machine Images (AMI): Templates with the OS and software used to launch instances.
• Elasticity and Auto Scaling: Capacity grows or shrinks automatically with demand.
• Storage options: EBS persistent volumes and instance store.
• Networking and security: VPC, security groups (virtual firewalls), key pairs, and Elastic IP addresses.
• Elastic Load Balancing across instances.
• Multiple regions and availability zones for high availability.
• Pricing models: On-demand, reserved, spot and savings plans.
• Monitoring through CloudWatch, and full control (root access) over the instance.
• Integration with other AWS services such as S3 and RDS.

Core components of Google App Engine

• Runtime environment (sandbox): The application runs in a secure sandbox in a language runtime (Python, Java, Go, Node.js, PHP). It has restrictions such as no arbitrary file writes or threads in the standard environment.
• Datastore (Firestore in Datastore mode): A scalable NoSQL schemaless object store with queries and transactions.
• Application servers and front-end load balancing: Requests are routed to instances, and the platform scales them automatically.
• Static file servers and CDN for static content.
• Services (APIs): Memcache, Task Queues, URL Fetch, Mail, Images, Users/Identity, Cron and Cloud Storage.
• Scheduler and scaling manager: Creates or shuts down instances and handles versions and traffic splitting.
• Application configuration (app.yaml) and the SDK/CLI for local testing and deployment.
• Admin console and monitoring/logging.

Windows Azure storage services (Azure Storage) provide durable, scalable, replicated cloud storage, accessed over REST APIs and SDKs. The main services are:

1. Blob storage: Stores unstructured data such as images, video, documents and backups. Types are block blobs (files), append blobs (logs) and page blobs (VM disks).
2. Table storage: A NoSQL key-attribute store for large amounts of structured, non-relational data.
3. Queue storage: Messaging between application components, supporting asynchronous, decoupled processing (for example from web roles to worker roles).
4. File storage: Fully managed shared file shares accessible through the SMB protocol.
5. Disk storage: Persistent disks for Azure virtual machines.

Common properties: Replication for durability (LRS, ZRS, GRS), high availability, scalability to petabytes, encryption, access keys/SAS tokens, and pay-per-use pricing. SQL Database is the separate relational service.

Business applications

• CRM and ERP (Salesforce, Microsoft Dynamics) run as SaaS without internal servers.
• Productivity and collaboration: Email, office suites, video conferencing.
• E-commerce and web hosting: Elastic resources handle seasonal peaks.
• Data storage, backup and disaster recovery.
• Business analytics and big data at low cost with pay-per-use.
• Development and testing environments on demand.
• Benefits: Lower capital cost, fast deployment, scalability, global access and better reliability.

Scientific applications

• High-performance computing: Large simulations (weather, physics, chemistry) use thousands of on-demand cores.
• Bioinformatics and genomics: Gene sequencing and analysis of huge data sets.
• Medical and healthcare: Remote monitoring and image processing.
• Astronomy and earth observation: Processing satellite and telescope data.
• Data sharing and collaboration among research groups.
• Benefits: Researchers get supercomputer-like capacity without owning it, pay only for experiments, scale quickly, and can use frameworks such as MapReduce.

A cloud computing service is a computing resource delivered over the Internet by a provider on demand, such as infrastructure, a development platform or a finished application. It is metered and billed by use, scales elastically, and the user does not manage the underlying hardware.

(a) Infrastructure-as-a-Service (IaaS)

• Provides virtual machines, storage, networks and load balancers on demand.
• The user installs and manages the OS, middleware and applications.
• Examples: Amazon EC2 and S3, Google Compute Engine, Azure Virtual Machines.
• Used for hosting websites, test environments, storage and backup, and high-performance computing.

(b) Platform-as-a-Service (PaaS)

• Provides a platform with runtime, middleware, database and development tools to build and deploy applications.
• The user controls only applications and data.
• Examples: Google App Engine, Heroku, Azure App Service, AWS Elastic Beanstalk.
• Used for fast application development and deployment.

(c) Software-as-a-Service (SaaS)

• A complete application is delivered through a web browser, usually on subscription. The provider handles everything.
• Examples: Gmail, Google Workspace, Microsoft 365, Salesforce, Dropbox, Zoom.
• Used for email, CRM, office tools and collaboration.

MapReduce is a programming model and framework (Google; implemented in Hadoop) for processing very large data sets in parallel on a cluster of commodity machines. The programmer writes two functions, and the framework handles splitting, scheduling, data movement, fault tolerance and load balancing.

• Map: $(k_1, v_1) \rightarrow \text{list}(k_2, v_2)$
• Reduce: $(k_2, \text{list}(v_2)) \rightarrow \text{list}(v_3)$

Example: word count on the input lines "cat dog cat" and "dog cat bird".

1. Input splitting: The input is divided into splits, one per map task (here line 1 and line 2).
2. Map function: For every word it emits (word, 1).
   • Mapper 1: (cat,1) (dog,1) (cat,1)
   • Mapper 2: (dog,1) (cat,1) (bird,1)
3. Shuffling: Intermediate pairs are transferred across the network so that all values with the same key go to the same reducer (partitioning by hash of the key).
4. Sorting: Each reducer sorts and groups the pairs by key.
   • bird -> [1], cat -> [1,1,1], dog -> [1,1]
5. Reduce function: It sums the list for each key.
   • Output: (bird,1) (cat,3) (dog,2)

Pseudo-code

map(key, line):
    for each word w in line:
        emit(w, 1)

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

An optional combiner performs a local reduce on the map side to reduce network traffic. Failed tasks are simply re-executed on another node.

Aneka is a .NET-based PaaS cloud application platform (by Manjrasoft) for building and deploying applications on public, private or hybrid clouds. It provides a flexible, service-oriented, container-based middleware that runs on clusters, desktops and virtual machines, with several programming models (thread, task, MapReduce).

Architecture The Aneka container is the basic unit deployed on every node. It hosts services and offers a uniform platform for them. The architecture has three layers of services:

1. Fabric services (lowest layer)

• Manage the physical and virtual resources of the infrastructure.
• Resource provisioning service (dynamic provisioning from public clouds such as EC2 or Azure).
• Hardware profiling and monitoring service (CPU, memory, disk information).
• Node and Platform Abstraction Layer (PAL) hide OS differences.

2. Foundation services (middle layer)

• Provide core cloud-infrastructure functions used by all applications.
• Membership and directory service, resource reservation, storage management, accounting and billing, licensing, monitoring, and security.

3. Application services (top layer)

• Support execution of applications for a given programming model.
• Scheduling services (job and task allocation) and execution services (running the work units). These are provided for the thread, task and MapReduce models.
• Also include data transfer and storage for applications.

Organisation and flow: Requests enter through the application layer, which schedules work and calls foundation services (security, accounting, reservation). These in turn use fabric services to get nodes and resources. The services are plugged into the container, so they can be configured per node as master, worker or both. An Aneka cloud is thus a network of containers coordinated by the master.

Properties of cloud computing (NIST essential characteristics and others)

• On-demand self-service: Users provision resources automatically without human help from the provider.
• Broad network access: Services are available over the network from many devices.
• Resource pooling (multi-tenancy): Provider resources are shared among many customers using virtualisation.
• Rapid elasticity: Resources scale up and down quickly, appearing unlimited.
• Measured service (pay-per-use): Usage is metered and billed, with transparency.
• Other properties: High availability and reliability, location independence, cost efficiency, automation, virtualisation, and service-oriented delivery.

Challenges in adapting to the cloud computing environment

• Security and privacy: Data breaches, shared infrastructure and loss of control over sensitive data.
• Compliance and legal issues: Data residency and regulations (GDPR, local laws).
• Vendor lock-in and interoperability: Moving between providers is difficult due to proprietary APIs.
• Downtime and reliability: Dependence on provider uptime and Internet connectivity.
• Performance and latency: Network delays for data-intensive applications.
• Cost management: Unexpected bills, hidden data-transfer charges.
• Migration complexity: Moving legacy applications and large data sets.
• Lack of skilled staff and resistance to change.
• Limited control and customisation, and unclear shared responsibility.
• Integration with existing on-premises systems, and SLA concerns.
• In developing countries such as Nepal, bandwidth and infrastructure limits also matter.

Hybrid cloud combines a private (or on-premises) cloud with a public cloud, with data and applications moving between them. Community cloud is shared by several organisations with common requirements.

Organisations adopt a hybrid cloud when:

• They have sensitive data or regulated workloads (banking records, health data) that must stay private, but also want cheap public resources for the rest.
• Demand has sudden peaks (festivals, exam results, sales) and they need cloud bursting into the public cloud.
• They are migrating gradually from on-premises systems and have existing investments to protect.
• They want disaster recovery, backup or development/test on the public cloud.
• They want to balance cost, control and flexibility, and avoid lock-in to one provider.

Organisations adopt a community cloud when:

• Several organisations have shared concerns such as security, compliance, policy or mission (government departments, universities, hospital networks, banks under one regulator).
• They want to share infrastructure cost while meeting common standards, with more privacy than a public cloud.
• They need to collaborate or share data and applications within the community, for example research consortia.
• A private cloud is too expensive for any single member.

Loosely coupled components are parts of an application that have minimal dependency on each other's internal details. They interact only through well-defined interfaces (APIs, web services, message queues) and can be developed, deployed, scaled and replaced independently.

How it is done

• Divide the application into independent services or modules (SOA or microservices).
• Communicate through standard protocols (REST, SOAP, JSON) or asynchronous messaging (queues such as Amazon SQS or Azure Queue).
• Avoid shared state, hard-coded addresses and tight links. Use service discovery and load balancers.
• Package components in VMs or containers.

Why it is done

• Scalability: Only the busy component is scaled, using elasticity.
• Fault tolerance: The failure of one component does not bring down the whole application, and retries are possible through queues.
• Flexibility and maintainability: Components can be updated or replaced without affecting others, in different languages.
• Reusability and faster development by independent teams.
• Better resource use and cost with pay-per-use.
• Portability and interoperability across clouds.

Virtualization creates a virtual (software-based) version of a computing resource so that one physical resource can be shared by many users or applications. The main types are:

1. Hardware (server) virtualization: A hypervisor divides one physical server into several isolated virtual machines (VMs), each with its own OS. It improves hardware utilisation and reduces cost. Examples: VMware ESXi, Xen, KVM, Hyper-V.

2. Operating-system-level virtualization: One OS kernel runs multiple isolated user-space instances (containers). It is lightweight and fast to start. Examples: Docker, LXC, OpenVZ.

3. Storage virtualization: Several physical storage devices are pooled and presented as a single logical storage unit (SAN, NAS, RAID, logical volumes). It simplifies management, backup and migration.

4. Network virtualization: Physical network resources (switches, routers, bandwidth) are combined into software-defined virtual networks such as VLANs, VPNs and SDN overlays.

5. Desktop virtualization: The user's desktop environment runs on a central server and is accessed remotely (VDI), allowing access from any device with central control and security.

6. Application virtualization: An application runs in an isolated environment without being installed on the client OS (for example Microsoft App-V, Citrix XenApp).

7. Memory and data virtualization: Physical memory of several machines is pooled (distributed shared memory), and data from different sources is presented through one logical view without caring where it is stored.

Cloud security threats arise because resources are shared, remotely accessed and controlled by a third party. The main issues are:

• Data breaches and data loss: Sensitive data can be stolen through weak access control, misconfiguration or insider attacks, or lost through accidental deletion, provider failure or lack of backup.
• Account and service hijacking: Stolen credentials or phishing give attackers control of cloud accounts and services.
• Insecure interfaces and APIs: Weak authentication or poorly designed APIs expose the cloud to attack.
• Denial of Service (DoS/DDoS): Attackers flood the service with traffic so that legitimate users cannot access it and resources are exhausted.
• Malicious insiders: Provider or customer employees misuse their privileged access.
• Multi-tenancy and shared-technology vulnerabilities: A flaw in the hypervisor or a side-channel attack can let one tenant reach another tenant's data.
• Insufficient due diligence and misconfiguration: Poorly configured storage, open ports or default settings expose resources.
• Vendor lock-in and loss of control: Dependence on a provider limits data portability and visibility into how security is handled.
• Compliance and legal issues: Data stored in other countries may violate regulations or be subject to foreign law.
• Malware injection and abuse of cloud services: Attackers upload malicious code or use cloud power for illegal activities such as password cracking.

Security monitoring in the cloud means continuously collecting and analysing data from cloud resources to detect attacks, misuse and policy violations in time.

Main activities

• Log collection and analysis: Logs from VMs, applications, firewalls, hypervisors and API calls (for example AWS CloudTrail, Azure Monitor) are gathered centrally and analysed.
• SIEM (Security Information and Event Management): Correlates events from many sources, raises alerts on suspicious patterns and supports investigation.
• Intrusion detection and prevention (IDS/IPS): Network and host-based sensors inspect traffic and system activity for known attack signatures and anomalies.
• Network traffic monitoring: Flow logs and packet analysis reveal unusual traffic such as DDoS or data exfiltration.
• Access and identity monitoring: Tracks logins, privilege changes and failed attempts to detect account compromise.
• Configuration and compliance monitoring: Checks continuously that resources follow security baselines and regulations, detecting misconfiguration.
• Vulnerability scanning: Regular scans of VMs, containers and applications find missing patches and weaknesses.
• Alerting and incident response: Automated alerts notify administrators, who contain and remediate the incident.

Because of shared responsibility, the provider monitors the underlying infrastructure while the customer monitors its own data, applications and access. Monitoring must be continuous, automated and capable of scaling with the cloud.

The main security mechanisms used to protect a cloud environment are:

• Encryption: Data is encrypted at rest and in transit (AES, TLS/SSL), protecting it from interception and unauthorised reading. Key management is essential.
• Authentication: Verifies identity using passwords, multi-factor authentication, digital certificates and tokens.
• Authorization and access control: Role-based access control (RBAC) and the principle of least privilege decide who can use which resource.
• Digital signatures and hashing: Ensure integrity and non-repudiation of data and messages.
• Single sign-on (SSO) and identity management: Centralised identity and federation across cloud services reduce credential risks.
• Firewalls and network security: Virtual firewalls, security groups, VPNs and network segmentation (VPC) restrict traffic.
• Intrusion detection and prevention systems: Detect and block malicious activity.
• Hardened virtual images and patch management: Secure VM images and regular updates reduce vulnerabilities.
• Isolation of tenants: Hypervisor-level separation of VMs, storage and networks protects multi-tenant systems.
• Auditing, logging and monitoring: Record activities for accountability and forensic analysis.
• Backup and disaster recovery: Protect against data loss and ensure availability.
• Cloud Access Security Brokers (CASB) and DDoS protection: Enforce security policy between users and cloud services and absorb floods of traffic.

Cloud computing gives scientists on-demand, scalable computing and storage without owning expensive clusters. Its main scientific applications are:

• High-performance and parallel computing: Large simulations in physics, chemistry, climate and weather modelling run on many cloud VMs, often through MapReduce or MPI-based clusters.
• Bioinformatics and genomics: Genome sequencing, protein folding and drug discovery need huge computation and storage, which the cloud provides (for example the 1000 Genomes data on AWS).
• Astronomy and space science: Telescope data (such as from the Square Kilometre Array) and satellite images are stored and processed on the cloud.
• Earth and environmental science: Satellite imagery, remote sensing, GIS and climate data analysis.
• Healthcare and medical research: Medical imaging analysis, epidemiology and large clinical data processing.
• High-energy physics: Experiments such as the Large Hadron Collider distribute data analysis over grid and cloud resources.
• Scientific workflows and collaboration: Researchers share data, tools and results and run reproducible workflows from anywhere.
• Big data analytics and machine learning: Training models and analysing huge datasets with Hadoop, Spark and GPU instances.

Benefits include pay-per-use cost, elasticity for burst workloads, easy data sharing and fast setup of experiments.

Cloud: Cloud computing is a model for delivering computing resources (servers, storage, databases, networking, software) as services over the Internet, on demand, with pay-per-use pricing, rapid elasticity, resource pooling and broad network access (NIST). Users do not own or manage the underlying hardware.

Types of clouds (deployment models)

1. Public cloud: Owned by a provider and offered to the general public over the Internet. Low cost, no maintenance, highly scalable. Used for web hosting, development/testing, backup and start-ups. Examples: AWS, Azure, Google Cloud.
2. Private cloud: Operated for a single organisation, on-premises or hosted. Gives greater control, security and compliance. Used by banks, governments and hospitals handling sensitive data. Examples: OpenStack, VMware private cloud.
3. Hybrid cloud: A combination of private and public clouds connected together. Sensitive workloads stay private while peak demand "bursts" to the public cloud. Used for cloud bursting, disaster recovery and gradual migration.
4. Community cloud: Shared by several organisations with common concerns (security, policy, compliance), such as universities or government agencies. Cost is shared and requirements are common.

(Service models are IaaS, PaaS and SaaS.)

Cloud service requirements

• On-demand self-service: users provision resources without provider interaction.
• Scalability and elasticity: resources grow or shrink with demand.
• High availability and reliability: fault tolerance, redundancy and SLAs.
• Security and privacy: encryption, access control, isolation of tenants.
• Performance: low latency and adequate bandwidth.
• Interoperability and portability: avoiding vendor lock-in through standards and APIs.
• Metering and billing: measured, pay-per-use service.
• Manageability and monitoring: dashboards, automation and resource tracking.
• Compliance and legal support: data location, auditing and regulatory requirements.
• Resource pooling and multi-tenancy: efficient sharing of resources among customers.

Significance of virtualization in cloud computing

Virtualization is the foundation of cloud computing because it allows physical resources to be abstracted, pooled and delivered on demand.

• Resource utilisation: many VMs share one physical server, raising utilisation and cutting hardware cost.
• Multi-tenancy and isolation: several customers run securely on the same hardware.
• Elasticity and scalability: VMs are created, resized or removed in minutes.
• Rapid provisioning: VM images and templates deploy quickly.
• Live migration and fault tolerance: VMs move between hosts without downtime and recover quickly after failure.
• Energy saving: fewer physical machines are needed (green computing).
• Hardware independence and portability: the same VM can run on different hardware.
• Testing and backup: snapshots and clones make backup, testing and disaster recovery easy.

Implementation levels of virtualization

1. Instruction Set Architecture (ISA) level: The guest ISA is emulated on a different host ISA by interpreting or dynamically translating instructions (for example QEMU, Bochs). It is flexible but slow.
2. Hardware abstraction level: A hypervisor (VMM) virtualizes CPU, memory and I/O directly over the hardware, giving high performance (Xen, KVM, VMware ESXi).
3. Operating system level: One OS kernel creates isolated containers (virtual environments) that share the kernel (Docker, LXC, OpenVZ). It has low overhead but all containers use the same OS type.
4. Library (user-level API) level: Applications are virtualized by intercepting the API calls to libraries (for example WINE, vCUDA).
5. Application level: An application runs on a process-level virtual machine such as the JVM or .NET CLR, giving portability.

Infrastructure requirements for virtualization

• Hardware: multi-core CPUs with virtualization support (Intel VT-x, AMD-V), large RAM, and enough storage and network capacity.
• Hypervisor / VMM software to create and manage VMs.
• Shared storage (SAN/NAS) to support live migration and VM images.
• High-speed, virtualized networking (virtual switches, VLANs, adequate bandwidth).
• Management software for provisioning, monitoring, orchestration and billing.
• Security tools: isolation, firewalls, access control.
• Reliable power, cooling and redundancy in the data centre, and licensed guest operating systems.

Task versus thread

MapReduce execution

MapReduce processes large data sets in parallel on a cluster. The input is divided into splits and the following phases run.

1. Map: Each map task reads a split and emits intermediate key-value pairs, $(k_1, v_1) \rightarrow list(k_2, v_2)$.
2. Shuffle: The intermediate pairs are partitioned by key (a hash function) and transferred across the network to the reducers, so that all values of the same key reach the same reducer.
3. Sort: Each reducer sorts and groups its input by key, giving $(k_2, list(v_2))$.
4. Reduce: The reduce function aggregates the list for every key and writes the output, $(k_2, list(v_2)) \rightarrow list(k_3, v_3)$.

A master (JobTracker) assigns tasks, monitors them and re-executes failed ones.

Example: word count for input lines "cloud is good" and "cloud is cheap".

• Map output: (cloud,1), (is,1), (good,1), (cloud,1), (is,1), (cheap,1)
• Shuffle and sort: cheap → [1]; cloud → [1,1]; good → [1]; is → [1,1]
• Reduce (sum the values): (cheap,1), (cloud,2), (good,1), (is,2)

Pseudo-code:

map(key, line):   for each word w in line: emit(w, 1)
reduce(word, counts): emit(word, sum(counts))

Cloud computing evolved gradually from earlier distributed computing models:

1. Mainframes and time-sharing (1950s-1960s): Expensive central computers were shared by many users through terminals. John McCarthy suggested computing as a public utility.
2. Virtualization (1960s-1970s): IBM's VM operating system allowed one machine to be divided into virtual machines.
3. Networking and the Internet (1970s-1990s): The ARPANET and the Web enabled remote access and client-server computing. Telecom companies offered VPNs.
4. Cluster computing (1980s-1990s): Many cheap computers were connected to form a powerful system.
5. Grid computing (1990s): Geographically distributed resources were coordinated for large scientific problems (Globus). It lacked easy commercial use.
6. Utility computing and ASPs (late 1990s): Computing was billed like electricity. Salesforce (1999) delivered applications over the Web (SaaS).
7. Modern cloud (2000s onwards): Amazon launched AWS (S3, EC2) in 2006; Google App Engine (2008), Microsoft Azure (2010) and OpenStack followed. Advances in virtualization, broadband, Web 2.0 and SOA made the cloud commercially practical.
8. Recent trends: containers and Kubernetes, serverless, edge and multi-cloud, and AI services.

SOA (Service-Oriented Architecture) is an architectural style in which software is built as a set of loosely coupled, reusable, interoperable services. Each service performs a well-defined business function, publishes its interface in a registry and is accessed over a network through standard protocols such as SOAP, REST, WSDL and XML/JSON. The parties are service provider, service consumer and service registry.

Why SOA suits cloud application development

• Loose coupling: services change independently, so applications are easy to update and scale.
• Reusability: the same service can be used by many applications, reducing development time and cost.
• Interoperability: standard protocols let services written in different languages and running on different platforms work together.
• Scalability and elasticity: each service can be scaled or deployed independently on cloud resources.
• Alignment with cloud models: IaaS, PaaS and SaaS are themselves delivered as services.
• Flexibility and composition: new applications can be assembled by orchestrating existing services (mashups, workflows).
• Easier integration of legacy systems and third-party cloud services.
• Maintainability and fault isolation: a failure in one service has limited impact.

A hypervisor (virtual machine monitor) is software that creates and runs virtual machines by allocating CPU, memory, storage and I/O of the physical host to each VM. There are two types.

Type 1: bare-metal (native) hypervisor

• Runs directly on the physical hardware without a host operating system.
• Offers high performance, strong isolation, security and scalability, so it is used in data centres and clouds.
• Examples: VMware ESXi, Microsoft Hyper-V, Citrix XenServer/Xen, KVM.

Type 2: hosted hypervisor

• Runs as an application on top of a conventional host operating system, which provides drivers and services.
• Easy to install and use, but has more overhead and lower performance and security because of the extra OS layer.
• Used for desktops, development and testing.
• Examples: Oracle VirtualBox, VMware Workstation, Parallels, QEMU.

By virtualization technique: full virtualization (guest OS unmodified, uses binary translation or hardware assist), paravirtualization (guest OS modified to call the hypervisor through hypercalls, as in Xen) and hardware-assisted virtualization (Intel VT-x, AMD-V).

Security architecture design in the cloud defines the layers, controls and responsibilities that protect cloud services, data and users. It follows the principle of defence in depth and the shared responsibility model: the provider secures the infrastructure while the customer secures its data, applications and access.

Main layers and components

1. Physical security: data-centre access control, surveillance, power and environmental protection.
2. Network security: firewalls, security groups, VPN, network segmentation (VPC), IDS/IPS and DDoS protection.
3. Host and virtualization security: hardened hypervisors and VM images, tenant isolation, patching and anti-malware.
4. Identity and access management: authentication, multi-factor authentication, SSO, RBAC and least privilege.
5. Data security: encryption at rest and in transit, key management, data masking, backup and secure deletion.
6. Application security: secure coding, API security, web application firewalls, vulnerability testing.
7. Security management and monitoring: logging, SIEM, auditing, incident response and compliance management.

Design principles: defence in depth, least privilege, separation of duties, secure defaults, continuous monitoring, redundancy for availability, and meeting the goals of confidentiality, integrity and availability (CIA). Standards and frameworks such as ISO 27001 and the Cloud Security Alliance guidance are followed.

The main security mechanisms implemented to secure a cloud are:

• Encryption: protects data at rest and in transit (AES, TLS/SSL, VPN) with proper key management.
• Authentication and multi-factor authentication: confirm user identity through passwords, tokens, biometrics and certificates.
• Authorization and access control: RBAC and least privilege limit what each user can do.
• Identity management and single sign-on: central identity, federation and lifecycle control of accounts.
• Digital signatures and hashing: guarantee integrity and non-repudiation.
• Firewalls, security groups and network segmentation: filter traffic and isolate workloads.
• IDS/IPS and DDoS protection: detect and stop attacks.
• Virtualization security: hypervisor hardening and tenant isolation.
• Logging, auditing and monitoring (SIEM): accountability and detection of abuse.
• Patch management and secure VM images: remove known vulnerabilities.
• Backup, replication and disaster recovery: keep data available after failure or attack.
• Data loss prevention and CASB: enforce policy on data leaving the cloud.
• Compliance and governance: SLAs, audits and standards (ISO 27001, SOC 2).

IaaS is the foundation layer on which PaaS is built, and SaaS (for example Gmail or Salesforce) sits above PaaS.

Aneka is a .NET-based Platform-as-a-Service framework (by Manjrasoft) for developing and deploying applications on private, public and hybrid clouds using multiple programming models (task, thread and MapReduce).

Aneka container: The container is the basic building block of an Aneka cloud. It is a lightweight runtime hosted on every node (physical or virtual) that provides the environment in which services are loaded and run. It hosts the services, handles communication between nodes and manages the node's resources. A cloud is formed by connecting many containers, and a node may act as a master or worker depending on the services it hosts.

Components of the Aneka framework (layers of the container)

1. Platform Abstraction Layer (PAL): hides differences among operating systems (Windows, Linux, Mono) and hardware, and gives uniform access to node information and resources.
2. Fabric services: include resource provisioning, hardware profiling and node membership/discovery, which manage the infrastructure.
3. Foundation services: provide the core functions: storage management, accounting and billing, resource reservation, licensing and authentication/security.
4. Application services: support the execution of applications and programming models: scheduling, execution, and the runtimes for task, thread and MapReduce models.
5. Management tools: the Aneka Management Console, SDK and APIs for configuring, monitoring and controlling the cloud.

Azure Shell (Azure Cloud Shell): An interactive, browser-based, authenticated shell provided in the Azure portal (and at shell.azure.com) for managing Azure resources. It offers both Bash and PowerShell, comes preinstalled with tools such as Azure CLI, Azure PowerShell, Git, kubectl, Terraform and editors, and uses an Azure Files share to keep files persistent. No local installation or setup is needed.

Azure monitoring services

• Azure Monitor: the central service that collects, analyses and acts on metrics and logs from applications, VMs, containers and networks. It provides dashboards, alerts and autoscale rules.
• Log Analytics: a workspace for querying logs with Kusto Query Language (KQL).
• Application Insights: application performance monitoring that tracks requests, response times, failures, dependencies and user behaviour.
• Azure Service Health and Resource Health: report outages, planned maintenance and the health of individual resources.
• Azure Advisor: gives recommendations on cost, performance, reliability and security.
• Azure Security Center / Microsoft Defender for Cloud: monitors security posture and detects threats.
• Network Watcher: monitors and diagnoses network issues.
• Alerts and Action Groups: notify by email, SMS or trigger automated responses.

Cloud platforms such as AWS, Azure and Google Cloud give businesses and consumers access to scalable computing, storage and software on a pay-per-use basis.

Business applications

• CRM and ERP: Salesforce, Microsoft Dynamics and SAP deliver customer, finance and supply-chain management as SaaS.
• Collaboration and productivity: Microsoft 365 and Google Workspace (email, documents, video meetings).
• Web and e-commerce hosting: scalable hosting that handles seasonal peaks.
• Data storage, backup and disaster recovery: low-cost offsite copies and fast recovery.
• Big data analytics and AI: processing large data sets, business intelligence and machine learning.
• Software development and testing: DevOps pipelines, test environments on demand.
• Human resources, accounting and payroll systems, plus IoT and mobile back-ends.

Consumer applications

• Email and storage: Gmail, Google Drive, Dropbox, iCloud.
• Social networking: Facebook, Instagram and similar.
• Media streaming: Netflix, YouTube, Spotify.
• Online office tools and photo storage: Google Docs, Google Photos.
• Online gaming and e-learning platforms.
• Mobile apps and online banking or shopping.

Benefits: lower cost, no maintenance, scalability, access from anywhere, automatic updates and quick deployment.