BSc CSIT (TU) Science Cloud Computing (BSc CSIT, CSC465) Question Paper 2077 Nepal

Scalability and Elasticity in Cloud Computing

Scalability

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

• Vertical scaling (scale up/down): Increasing the power of an existing machine (more CPU, RAM). Limited by hardware ceiling; usually requires downtime.
• Horizontal scaling (scale out/in): Adding or removing more machines/instances behind a load balancer. Preferred in the cloud because it offers near-unlimited growth and fault tolerance.

Elasticity

Elasticity is the ability to automatically and dynamically acquire and release resources in real time to match the current workload, so the system uses only what it needs at any instant. Scalability is the capacity; elasticity is the automatic, on-demand exercise of that capacity. Example: an e-commerce site automatically adds servers during a festival sale and removes them afterwards (e.g. AWS Auto Scaling Groups).

Load Balancing Techniques

A load balancer distributes incoming traffic across multiple backend servers to maximize throughput, minimize response time and avoid overload. Common algorithms:

Examples: AWS Elastic Load Balancer (ALB/NLB), Google Cloud Load Balancing, Nginx/HAProxy.

Auto-Scaling Mechanisms

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

• Reactive / metric-based scaling: Scales when a monitored metric (CPU > 70%, request count, queue length) crosses a threshold, e.g. via AWS CloudWatch alarms.
• Scheduled scaling: Adds capacity at known peak times (e.g. 9 AM office login surge).
• Predictive scaling: Uses ML on historical data to provision capacity before demand arrives.

Example flow: CloudWatch detects average CPU > 70% across an Auto Scaling Group → triggers a scale-out policy → launches 2 new EC2 instances → ELB starts routing traffic to them. When load drops, instances are terminated (scale-in), keeping cost low while meeting demand.

Cloud Computing

Cloud computing is a model for enabling ubiquitous, convenient, on-demand network access to a shared pool of configurable computing resources (networks, servers, storage, applications, services) that can be rapidly provisioned and released with minimal management effort (NIST definition). Users pay only for what they use (pay-as-you-go).

Essential Characteristics (NIST)

1. On-demand self-service – users provision resources automatically without human interaction with the provider.
2. Broad network access – services available over the network through standard devices (laptops, phones).
3. Resource pooling – provider resources are pooled and dynamically assigned (multi-tenancy).
4. Rapid elasticity – resources scale out and in quickly to match demand.
5. Measured service – usage is metered, monitored and billed (pay-per-use).

Service Models

Moving from IaaS → PaaS → SaaS, the user's responsibility decreases and abstraction increases.

Deployment Models

• Public cloud: Owned by a provider and shared by the public over the internet (e.g. AWS, Azure). Low cost, high scalability, less control.
• Private cloud: Dedicated to a single organization, on-premises or hosted. High control and security; higher cost (e.g. OpenStack private cloud).
• Hybrid cloud: Combines public and private clouds, allowing data/apps to move between them (e.g. keep sensitive data on-premises, burst to public cloud during peak load).
• Community cloud: Shared by several organizations with common concerns (security, compliance), e.g. a cloud shared by government agencies or banks.

Virtualization

Virtualization is the technology of creating a software-based (virtual) version of physical resources such as servers, storage, networks or operating systems. A single physical machine is partitioned into multiple isolated virtual machines (VMs), each running its own OS and applications, managed by a hypervisor.

Role of Virtualization in Cloud Computing

Virtualization is the foundation of cloud computing because it enables:

• Resource pooling and multi-tenancy – many customers share the same physical hardware safely and in isolation.
• Higher utilization & cost efficiency – idle physical capacity is consolidated.
• Rapid provisioning and elasticity – VMs can be created, cloned and destroyed in seconds.
• Isolation and fault containment – a crash in one VM does not affect others.
• Portability & live migration – VMs can be moved between hosts for load balancing and maintenance.

Types of Virtualization

1. Server (hardware) virtualization – multiple VMs on one server.
2. Storage virtualization – pooling physical storage into a single logical unit (e.g. SAN).
3. Network virtualization – software-defined networks, VLANs, virtual switches.
4. Desktop virtualization – VDI; desktops hosted centrally and accessed remotely.
5. Application/OS-level virtualization – containers (e.g. Docker) share the host kernel.

Hypervisor and its Types

A hypervisor (Virtual Machine Monitor) is the software layer that creates, runs and manages VMs and allocates physical resources to them.

Type 1 – Bare-metal hypervisor

• Runs directly on the physical hardware (no host OS).
• Higher performance, better security, used in data centers/cloud.
• Examples: VMware ESXi, Microsoft Hyper-V, Xen, KVM.

Type 2 – Hosted hypervisor

• Runs as an application on top of a host operating system.
• Easier to install, but extra overhead and lower performance; used for desktops/testing.
• Examples: VMware Workstation, Oracle VirtualBox.

Diagram (in words): In Type 1 the stack is Hardware → Hypervisor → VMs, whereas in Type 2 it is Hardware → Host OS → Hypervisor → VMs.

The essential characteristics of cloud computing (NIST) are:

1. On-demand self-service – Consumers can provision computing resources (server time, storage) automatically, without requiring human interaction with the provider.
2. Broad network access – Capabilities are available over the network and accessed through standard mechanisms by heterogeneous client devices (laptops, mobiles, tablets).
3. Resource pooling – The provider's resources are pooled to serve multiple consumers using a multi-tenant model, with resources dynamically assigned and reassigned on demand.
4. Rapid elasticity – Resources can be elastically provisioned and released, scaling out and in quickly to match demand, appearing unlimited to the consumer.
5. Measured service – Resource usage is automatically controlled, metered and monitored, providing transparency and pay-per-use billing.

IaaS vs PaaS vs SaaS

Summary: As we move IaaS → PaaS → SaaS, the level of abstraction increases and the management burden on the user decreases. IaaS rents infrastructure, PaaS rents a platform to build apps, and SaaS delivers finished software over the internet.

Virtualization

Virtualization is the creation of a software-based (virtual) version of a physical resource such as a server, storage device, network or operating system. Using a hypervisor, one physical machine is divided into multiple isolated virtual machines, each with its own OS and applications.

Role in Cloud Computing

Virtualization is the backbone of cloud computing because it enables:

• Resource pooling and multi-tenancy – multiple users securely share the same physical hardware in isolation.
• Efficient utilization – consolidating workloads raises hardware usage and lowers cost.
• Rapid, elastic provisioning – VMs/containers can be spun up or torn down in seconds to match demand.
• Isolation & security – failures or breaches in one VM do not affect others.
• Portability & live migration – VMs can move across hosts for load balancing, high availability and maintenance.

Without virtualization, the on-demand, scalable, pay-per-use model of the cloud would not be practical.

Public vs Private vs Hybrid Cloud

Summary:

• Public cloud – resources delivered over the internet to many tenants; cost-effective and highly scalable but less control.
• Private cloud – infrastructure dedicated to one organization; greater security and control but higher cost.
• Hybrid cloud – integrates both, keeping critical/sensitive workloads in the private cloud while using the public cloud for scalability and burst capacity (cloud bursting).

MapReduce Programming Model

MapReduce is a programming model (popularized by Google, implemented in Hadoop) for processing very large datasets in parallel across a distributed cluster. Computation is expressed as two functions:

1. Map – takes input key-value pairs and produces a set of intermediate key-value pairs.
2. Reduce – merges all intermediate values associated with the same key to produce the final output.

Between them, the framework performs an automatic shuffle and sort step that groups intermediate pairs by key. The framework also handles parallelization, data distribution, fault tolerance and scheduling.

Example: Word Count

Count the occurrences of each word in a large document.

Input:  "cloud is cloud"

Map:    (cloud,1) (is,1) (cloud,1)

Shuffle/Sort (group by key):
        cloud -> [1,1]
        is    -> [1]

Reduce: (cloud, 2) (is, 1)

Pseudocode:

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

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

The map phase runs in parallel on data splits across many nodes, and the reduce phase aggregates results, allowing terabytes of data to be processed efficiently.

Service Level Agreement (SLA)

A Service Level Agreement (SLA) is a formal contract between a cloud service provider and a customer that defines the level of service expected, the metrics by which it is measured, and the penalties or remedies if the agreed levels are not met. It sets clear, measurable expectations and protects both parties.

Key Components

1. Service description – the services provided and their scope.
2. Performance / service-level objectives (SLOs) – measurable targets such as uptime/availability (e.g. 99.9%), response time, throughput.
3. Metrics and measurement method – how performance is measured and reported.
4. Responsibilities – duties of both provider and customer.
5. Penalties / service credits – compensation (e.g. credits) when SLAs are breached.
6. Security and compliance terms – data protection, privacy, regulatory obligations.
7. Support and escalation procedures – help-desk availability, incident handling.
8. Termination and review clauses – conditions for exit and periodic review.

Major Security Issues in Cloud Computing

1. Data breaches and data loss – Sensitive data stored on shared infrastructure may be exposed by attacks, misconfiguration, or accidental deletion with no backup.
2. Data privacy and location – Data may be stored across multiple jurisdictions, raising compliance (e.g. GDPR) and confidentiality concerns.
3. Multi-tenancy / isolation failure – Because tenants share hardware, a flaw in isolation can let one tenant access another's data (side-channel attacks).
4. Insecure interfaces and APIs – Weak or exposed management APIs can be exploited to gain unauthorized access.
5. Account / credential hijacking – Stolen credentials let attackers control cloud resources; weak authentication increases risk.
6. Insider threats – Malicious or careless provider/customer staff with privileged access.
7. Denial of Service (DoS/DDoS) – Attacks that overwhelm cloud services and degrade availability.
8. Loss of control / vendor lock-in – Customers depend on the provider for security controls and may find it hard to migrate.

Mitigations: encryption (at rest and in transit), strong IAM and multi-factor authentication, regular auditing, and clear SLAs defining security responsibilities (shared responsibility model).

Hypervisor

A hypervisor (Virtual Machine Monitor, VMM) is a software layer that creates and runs virtual machines (VMs). It abstracts and allocates the physical hardware (CPU, memory, storage, I/O) among multiple VMs, allowing several isolated guest operating systems to run on a single physical host simultaneously.

Type 1 vs Type 2 Hypervisors

Summary: Type 1 hypervisors run natively on hardware for maximum performance and are used in production cloud environments, while Type 2 hypervisors run as an application within a host OS and are convenient for personal/testing use.

Hadoop Distributed File System (HDFS)

HDFS is the primary storage system of Hadoop, designed to reliably store very large files across a cluster of commodity machines and to support high-throughput, batch processing of big data (used with MapReduce).

Architecture (Master–Slave)

• NameNode (Master): Manages the file-system namespace and metadata (file names, directory structure, block locations). Single point of coordination; a Secondary/Standby NameNode aids recovery.
• DataNodes (Slaves): Store the actual data blocks and serve read/write requests from clients.

Key Features

• Block storage: Files are split into large blocks (default 128 MB) distributed across DataNodes.
• Replication for fault tolerance: Each block is replicated (default factor 3) on different nodes, so data survives node failures.
• Write-once, read-many model: Optimized for batch processing rather than frequent updates.
• Data locality: Computation is moved to the node holding the data, reducing network traffic.
• Scalability: Runs on inexpensive commodity hardware and scales to thousands of nodes/petabytes.
• High throughput: Prioritizes streaming access to large datasets over low-latency random access.

Summary: HDFS provides reliable, fault-tolerant, scalable storage for big data by splitting files into replicated blocks managed by a NameNode and stored on many DataNodes.