GTU 3170717 Cloud Computing Winter 2025 Solved Paper & Solutions

PaaS (Platform as a Service) is a cloud service that gives developers a ready platform to build, test and run applications. The provider manages the servers, the operating system, the runtime and the database. The user manages only the application and its data. Examples: Google App Engine and Azure App Service.

A public cloud suits normal workloads and fast growth. A private cloud suits sensitive data and strict rules.

Virtualization is the technique of creating a virtual version of a physical resource, such as a server, a storage device or a network, so that one physical resource can act as many separate ones. A software layer called the hypervisor sits between the hardware and the operating systems and shares the hardware among virtual machines (VMs).

Types of virtualization

• Server virtualization: one physical server runs many VMs.
• Storage virtualization: many storage devices appear as one pool.
• Network virtualization: one physical network is divided into many virtual networks.
• Desktop virtualization: a user desktop runs on a central server.

Benefits

• Better use of hardware, since many VMs share one machine.
• Each VM is isolated from the others.
• A VM can be created, copied or moved quickly.
• It lowers the cost of hardware, space and power.
• It gives the base for cloud services, because a cloud provider can give each customer a separate VM on shared hardware.

Application: server consolidation

A company has 10 servers, and each one uses only 10 to 15 percent of its capacity. With virtualization, the 10 applications run as 10 VMs on one or two powerful servers. The hardware, power and maintenance cost falls, and the cloud providers use the same idea to rent a VM to each customer.

The three basic components of the cloud architecture are:

1. Clients: the devices that the users use, such as laptops, phones and browsers.
2. Data centre: the collection of servers that hold the applications and the data.
3. Distributed servers: servers at different places, which give reliability and speed.

The three services given to the user are IaaS, PaaS and SaaS.

Virtualization is the base of cloud computing, and it adds the following new dimensions.

• Resource pooling and multi-tenancy: many customers share the same hardware, safely isolated from each other.
• Elasticity: VMs can be added or removed in minutes, as the load changes.
• Rapid provisioning: a new server is a VM image started in seconds, not a machine bought over weeks.
• Hardware independence: a VM can run on any physical host, and it can be moved with live migration.
• Cost efficiency: high use of hardware lowers the price for each customer.
• Disaster recovery: VM copies can be restored on another host easily.

So the cloud model of pay-as-you-use and on-demand service is possible because of virtualization.

Microsoft Azure is the cloud platform of Microsoft. It gives hundreds of services. The main groups are shown below.

Short notes on the main services

• Virtual Machines give Windows or Linux servers on demand. This is the IaaS service of Azure.
• App Service hosts web and mobile applications without managing servers. This is a PaaS service.
• Azure Functions run small pieces of code when an event occurs, and the user pays only for the run time. This is the serverless model.
• Blob storage keeps large amounts of unstructured data such as images and videos.
• Azure SQL Database is a managed relational database.
• Azure Active Directory manages user identity and sign-in.

Azure supports IaaS, PaaS and SaaS, and also hybrid cloud, so a company can join its own data centre to Azure. The services are spread over many regions of the world, and the user pays only for what is used.

The main challenges for cloud security are:

1. Data breaches and data loss.
2. Account hijacking through stolen passwords.
3. Insecure interfaces and APIs.
4. Multi-tenancy: weak isolation between customers who share hardware.
5. Insider threats from staff of the provider or the customer.
6. Denial of service attacks.
7. Compliance and legal issues about where the data is kept.
8. Lack of control and visibility over the infrastructure.

Eucalyptus is open-source software for building a private or hybrid cloud inside an organisation's own data centre. The name stands for Elastic Utility Computing Architecture for Linking Your Programs To Useful Systems. It was developed at the University of California, Santa Barbara.

Its main feature is that its API is compatible with Amazon Web Services, so tools written for AWS (EC2 and S3) also work with it.

Components

• Cloud Controller (CLC): the front end. It handles the user requests and manages the cloud.
• Walrus: the storage service, similar to Amazon S3.
• Cluster Controller (CC): manages a group of nodes and the network of that group.
• Storage Controller (SC): provides block storage to the VMs, similar to Amazon EBS.
• Node Controller (NC): runs on each machine and controls the VMs on it.

A hypervisor, or virtual machine monitor, is the software that creates and runs virtual machines. Virtualization in the cloud depends on it, since every VM that a customer rents is run and controlled by a hypervisor. It shares the CPU, memory, storage and network of the physical host among the VMs and keeps the VMs isolated from each other.

Type 1: bare-metal hypervisor

It runs directly on the hardware, with no host operating system below it. It has direct control of the hardware, so it is fast and secure. The VMs on it are called guests. It is used in data centres and clouds.

Examples: VMware ESXi, Microsoft Hyper-V and Xen.

Type 2: hosted hypervisor

It runs as an application on a host operating system, such as Windows or Linux. The host OS gives it access to the hardware. It is easy to install, but slower, because every request goes through the host OS. It is used on desktops for learning and testing. A fault in the host OS can affect all the VMs on it, so it is not used for large production servers.

Examples: Oracle VirtualBox and VMware Workstation.

Comparison

The two types are compared below.

Load balancing is the distribution of the incoming work, such as user requests, evenly among many servers or VMs, so that no single one is overloaded.

• A load balancer receives the requests and forwards each one to a suitable server.
• Common rules are round robin, least connections and weighted distribution.
• It improves speed and availability, because a failed server is skipped.
• Example: the AWS Elastic Load Balancer.

A Storage Area Network (SAN) is a dedicated high-speed network that gives servers block-level access to shared storage devices, using Fibre Channel or iSCSI.

Autonomic computing means that a system manages itself, with little human help. Its four properties are self-configuring, self-healing, self-optimising and self-protecting.

An autonomic security SAN applies the self-protecting idea to the storage network. The security is managed by the system itself, with policies, and not by hand.

• It monitors the SAN for attacks and unusual access.
• It applies the security rules automatically, such as zoning, LUN masking, access control and encryption of the data.
• It responds to a threat by itself, for example by blocking a device that behaves badly.
• It protects the stored data of the cloud and lowers the work of the administrators.

Amazon Glacier (now Amazon S3 Glacier) is a very low-cost cloud storage service for archives and long-term backups, which are rarely read. Its data is stored with very high durability, designed for 99.999999999 percent (eleven nines).

How it works

1. The user creates a vault, a container for archives.
2. Files are uploaded as archives. Each archive gets a unique ID.
3. To read the data, the user starts a retrieval job and waits. The data is not available instantly.
4. When the job is complete, the user downloads the data from the output of the job.

The retrieval time depends on the option chosen: expedited takes about 1 to 5 minutes, standard takes about 3 to 5 hours, and bulk takes about 5 to 12 hours. A faster option costs more. Glacier suits data such as old records, medical images, logs and legal documents that must be kept for years, but that are almost never read. The user should not use it for data that needs quick access.

Glacier compared with S3

A virtual cluster is a group of virtual machines that are connected by a virtual network and work together like a physical cluster. The VMs may run on different physical servers.

• Its size can change: VMs are added or removed as the load changes.
• Each virtual cluster can have its own operating system and software, and it is isolated from other clusters.
• A VM can be moved from one host to another without stopping the cluster.

A cloud has a compute cloud that runs the applications and a storage cloud that holds the data. The design is in layers.

• Cloud gateway: the entry point of the cloud. It accepts the user requests and checks the identity.
• Resource manager and scheduler: decides which VM runs a job and where the data goes. It also monitors the load and the billing.
• Compute cloud: a large cluster of servers that run VMs created from images.
• Storage cloud: a distributed file system that spreads the data over many disks and keeps copies, for reliability.

Security and monitoring work across all the layers, and the same design lets the cloud add more servers or disks when the demand grows.

Disaster recovery (DR) in the cloud is the set of plans and tools that bring the systems and the data back after a failure or a disaster, such as a fire, a flood, a cyber attack or a power failure, so that the business continues.

Key measures

• Recovery Time Objective (RTO): the longest time the service may be down.
• Recovery Point Objective (RPO): the largest amount of data loss, measured in time, that is acceptable.

Steps to perform disaster recovery

1. Risk assessment: find the critical systems and the threats to them.
2. Set the RTO and the RPO for each system.
3. Take backups of the data and of the VM images regularly.
4. Replicate the data and the systems to another availability zone or region, away from the main site.
5. Automate the failover, so the traffic moves to the standby site when the main site fails.
6. Test the plan regularly with drills, and update it.
7. Failback: return to the main site after it is repaired.

Common strategies

The cloud makes DR cheaper, because the standby resources are paid for only when used, and the data can be stored in many regions.