BSc CSIT (TU) Science Cloud Computing (BSc CSIT, CSC465) Question Paper 2079 Nepal

Layered Architecture of Cloud Computing

Cloud computing is organized as a stack of layers, where each higher layer is built on the services exposed by the layer below it.

Each layer hides the complexity of the one below and offers a clean interface, supporting the separation of concerns and pay-per-use model.

Cloud Broker

A cloud broker is an intermediary entity that manages the use, performance and delivery of cloud services and negotiates relationships between cloud providers and cloud consumers. Its services fall into three categories:

• Service Intermediation – enhances a given service (e.g. extra security, identity management).
• Service Aggregation – combines multiple services into one and ensures data integration between them.
• Service Arbitrage – like aggregation, but the services being combined are flexible/dynamic (the broker chooses the best provider by price or quality).

Cloud Carrier

A cloud carrier is the intermediary that provides connectivity and transport of cloud services from the provider to the consumer, typically through telecommunication and network providers. It ensures the network access (Internet/dedicated links) meets the SLA, providing the secure and reliable communication channel between the consumer and the provider.

NIST Cloud Reference Model

The NIST cloud reference architecture defines five major actors:

1. Cloud Consumer – the person/organization that uses the service.
2. Cloud Provider – makes the service available (service deployment, orchestration, management, security, privacy).
3. Cloud Auditor – independently assesses services, security and performance.
4. Cloud Broker – manages and negotiates services between consumers and providers.
5. Cloud Carrier – provides connectivity/transport of services.

This reference model standardizes the roles, responsibilities and interactions among the actors so that cloud ecosystems can interoperate.

Scalability vs Elasticity

Scalability is the ability of a system to handle growing workload by adding resources. It is usually a planned, longer-term capability.

• Vertical scaling (scale-up): increase the capacity of a single node (more CPU/RAM).
• Horizontal scaling (scale-out): add more nodes/instances to share the load.

Elasticity is the ability to automatically and dynamically add or remove resources in real time to match the current demand, so the system uses only what it needs. Elasticity is short-term and automatic, whereas scalability is the broader capacity to grow. Example: an e-commerce site automatically spins up extra VMs during a festival sale and releases them afterward — that is elasticity built on a scalable architecture.

Load Balancing Techniques

A load balancer distributes incoming requests across multiple servers to avoid overload, maximize throughput and ensure high availability. Common algorithms:

• Round Robin – requests sent to each server in turn.
• Weighted Round Robin – stronger servers get a larger share.
• Least Connections – new request goes to the server with the fewest active connections.
• Least Response Time – goes to the fastest-responding server.
• IP Hash – server chosen by hashing the client IP (gives session stickiness).

Example: AWS Elastic Load Balancer (ELB) spreads HTTP traffic across EC2 instances in multiple availability zones.

Auto-Scaling Mechanisms

Auto-scaling automatically adjusts the number of running instances based on monitored metrics.

• Reactive / Dynamic scaling: scales based on real-time metrics such as CPU > 70% (e.g. if average CPU > 70% for 5 min, add 1 instance; if < 30%, remove 1).
• Scheduled scaling: scales at known times (e.g. add capacity every weekday 9 AM).
• Predictive scaling: uses ML/historical patterns to provision in advance.

Key components: a scaling policy, a minimum/maximum/desired count, and health checks that replace unhealthy instances. Example: an AWS Auto Scaling Group behind an ELB grows from 2 to 10 EC2 instances when traffic spikes and shrinks back when it subsides, combining elasticity, load balancing and high availability.

What is Cloud Computing?

Per the NIST definition, cloud computing is a model for enabling ubiquitous, convenient, on-demand network access to a shared pool of configurable computing resources (e.g. networks, servers, storage, applications and services) that can be rapidly provisioned and released with minimal management effort or service-provider interaction.

Essential Characteristics (NIST)

1. On-demand self-service – a consumer can provision resources automatically without human interaction.
2. Broad network access – capabilities are available over the network and accessed through standard mechanisms (laptops, phones, tablets).
3. Resource pooling – provider resources are pooled to serve multiple tenants using a multi-tenant model (location independence).
4. Rapid elasticity – capabilities can be elastically provisioned and released, appearing unlimited to the consumer.
5. Measured service – resource use is metered (pay-per-use) and can be monitored, controlled and reported.

Service Models

Deployment Models

• Public Cloud – infrastructure owned by a provider and offered to the general public over the Internet. Example: AWS, Azure. (Low cost, shared.)
• Private Cloud – infrastructure operated solely for a single organization (on-premises or hosted). Example: a bank's internal cloud. (More control/security.)
• Hybrid Cloud – a combination of public and private clouds bound together, allowing data/app portability (e.g. keep sensitive data private, burst to public for peak loads).
• Community Cloud – shared by several organizations with common concerns (security, compliance). Example: a cloud shared by government health agencies.

Conclusion

Cloud computing delivers IT as a utility through these five characteristics, three service models and four deployment models, offering cost savings, scalability and flexibility.

The five essential characteristics of cloud computing (NIST) are:

1. On-demand self-service – users provision computing resources (server time, storage) automatically, without requiring human interaction with the provider.
2. Broad network access – services are available over the network and accessed via standard devices (PCs, mobiles, tablets).
3. Resource pooling – the provider's resources are pooled to serve multiple consumers using a multi-tenant model; specific physical location is abstracted.
4. Rapid elasticity – resources can be scaled out and in quickly (often automatically) to match demand, appearing virtually unlimited.
5. Measured service – resource usage is metered, monitored and reported, enabling a transparent pay-per-use billing model.

IaaS vs PaaS vs SaaS

Summary: IaaS gives the most control and flexibility (you manage the OS upward); PaaS abstracts the infrastructure so developers focus only on code; SaaS delivers finished software requiring no management by the user. Control decreases and convenience increases as you move IaaS → PaaS → SaaS.

Virtualization

Virtualization is the technique of creating a virtual (software-based) version of a physical resource — such as a server, storage device, network or operating system — so that a single physical machine can be partitioned into multiple isolated virtual machines (VMs), each running its own OS and applications. It is achieved through a software layer called a hypervisor that sits between the hardware and the VMs.

Role in Cloud Computing

Virtualization is the foundational technology of cloud computing because it enables:

• Resource pooling & multi-tenancy – many customers share the same physical hardware while staying isolated.
• Elasticity & rapid provisioning – VMs can be created, cloned or destroyed within minutes to match demand.
• Higher utilization & cost efficiency – idle physical capacity is consolidated, reducing hardware and power costs.
• Isolation & security – a failure or breach in one VM does not affect others.
• Portability & migration – VMs can be moved (live-migrated) across hosts for load balancing and fault tolerance.

Without virtualization, the on-demand, pay-per-use and scalable nature of cloud computing would not be feasible.

Public vs Private vs Hybrid Cloud

Summary: A public cloud is cost-effective and highly scalable but offers less control; a private cloud offers maximum security and control at higher cost; a hybrid cloud combines both, keeping sensitive workloads on the private side while using the public cloud for scalability and peak demand (cloud bursting).

MapReduce Programming Model

MapReduce is a programming model for processing and generating large data sets in parallel across a distributed cluster. A computation is expressed as two user-defined functions:

• Map(key, value) → list(key', value') – processes input records and emits intermediate key-value pairs.
• Reduce(key', list(value')) → list(value'') – aggregates all values sharing the same intermediate key.

Between them, the framework performs a Shuffle & Sort phase that groups all intermediate pairs by key before sending them to reducers. The framework handles parallelization, data distribution, fault tolerance and load balancing automatically.

Example: Word Count

Count the occurrences of each word in a large text corpus.

Map(docId, line):
    for each word w in line:
        emit(w, 1)

// Shuffle & Sort groups by word:
// "cloud" -> [1, 1, 1],  "data" -> [1, 1]

Reduce(word, counts):
    sum = 0
    for c in counts:
        sum += c
    emit(word, sum)

If the input is "cloud data cloud", the Map phase emits (cloud,1),(data,1),(cloud,1); after shuffle the Reduce phase outputs (cloud,2),(data,1). This model underlies frameworks like Apache Hadoop.

Service Level Agreement (SLA)

A Service Level Agreement (SLA) is a formal, often contractual, document between a cloud service provider and a cloud consumer that defines the expected level of service, the metrics by which it is measured, the responsibilities of each party, and the penalties/remedies if the agreed levels are not met.

Key Components

1. Service description – the scope of services covered.
2. Performance metrics / SLOs – measurable targets such as availability/uptime (e.g. 99.9%), response time, throughput.
3. Monitoring and reporting – how performance is measured and reported.
4. Responsibilities – duties of both provider and consumer.
5. Penalties / remedies – credits or compensation when SLAs are violated.
6. Security and compliance – data protection, privacy and regulatory obligations.
7. Exclusions and exceptions – conditions not covered (e.g. scheduled maintenance, force majeure).
8. Termination and renewal terms – duration, exit conditions and review process.

Major Security Issues in Cloud Computing

1. Data breaches & data loss – sensitive data stored off-premises may be exposed through attacks, or lost due to provider failure with no backup.
2. Data privacy & confidentiality – multi-tenancy and storing data across jurisdictions raise concerns over who can access the data and which laws apply (data sovereignty).
3. Insecure interfaces and APIs – weak or poorly designed management APIs can be exploited by attackers.
4. Account / credential hijacking – stolen credentials let attackers eavesdrop, manipulate data or redirect traffic.
5. Insider threats – malicious or careless provider/customer staff with privileged access.
6. Multi-tenancy / isolation failure – flaws in virtualization (VM escape, side-channel attacks) may let one tenant access another's data.
7. Denial of Service (DoS/DDoS) – overwhelming services to make them unavailable.
8. Loss of control & vendor lock-in – limited visibility into provider's security and difficulty migrating away.
9. Compliance and legal issues – meeting regulations such as GDPR, HIPAA across shared infrastructure.

Mitigations include strong encryption (in transit and at rest), robust IAM and multi-factor authentication, clear SLAs, auditing, and well-defined isolation mechanisms.

Hypervisor

A hypervisor (Virtual Machine Monitor, VMM) is the software/firmware layer that creates, runs and manages virtual machines by abstracting the physical hardware and allocating CPU, memory and I/O resources among multiple guest operating systems, keeping them isolated from one another. It is the core enabler of virtualization in cloud computing.

Type 1 vs Type 2 Hypervisors

Summary: A Type 1 hypervisor runs bare-metal directly on hardware giving better performance and security (used in clouds/data centers), while a Type 2 hypervisor runs as an application on an existing OS and is suited to personal use and testing.

Hadoop Distributed File System (HDFS)

HDFS is the primary storage system of Apache Hadoop, designed to store very large files reliably across a cluster of commodity hardware and to stream that data to applications at high throughput. It follows a write-once, read-many access model and is optimized for batch processing.

Architecture (Master–Slave)

• NameNode (Master) – stores the filesystem metadata (namespace, directory tree, mapping of blocks to DataNodes). A single logical NameNode manages the namespace.
• DataNodes (Slaves) – store the actual data blocks and serve read/write requests; they periodically send heartbeats and block reports to the NameNode.
• Secondary NameNode – periodically merges the edit log with the filesystem image (checkpointing); it is not a hot standby.

Key Features

• Block storage: files are split into large blocks (default 128 MB) distributed across DataNodes.
• Replication: each block is replicated (default factor 3) for fault tolerance; if a DataNode fails, replicas on other nodes are used and re-replicated.
• Fault tolerance & high availability through replication and heartbeats.
• Scalability: simply add commodity DataNodes to grow capacity.
• Data locality: computation (MapReduce) is moved to where the data resides, reducing network traffic.

HDFS thus provides reliable, scalable and cost-effective storage for big-data workloads.