Abasyn University

Department of Computing | Islamabad Campus

Official Course Code: CS445

Parallel & Distributed Computing

Comprehensive Quick Preparation Notes (Final Term Exam)

Side-by-Side Dual Format: Professional English + Simple Roman English

Course Lecturer Ms. Nadra Bibi

Exam Date & Time Monday, 22nd May 2026 (09:00 AM - 11:00 AM)

Exam Format 5 Core Scenario-Based Questions (10 Marks Each = Total 50)

Content Source Official CS445 PPT Slides & Curriculum Blueprint

This document contains all high-yield concepts mapped strictly out of the required curriculum blocks.

CS445: Distributed & Parallel Computing - Official Exam Notes Page 1 of 6

 Topic 1: Amazon Web Services (AWS Architecture)

PROFESSIONAL ENGLISH EASY ROMAN ENGLISH

1. Amazon Route 53 (Global Scalable DNS): 1. Amazon Route 53 (DNS Domain System):
• Acts as a highly available Domain Name System • Ye ek smart online phonebook ki tarah hai. Jab koi
(DNS) that safely translates human-readable user website ka naam type karta hai, to Route 53
domain URLs (e.g., website names) into computer- usko computer ki numeric language (IP Address)
understandable IP addresses. me change kar deta hai.
• Handles millions of concurrent queries via global • Jab lakhon log ek sath aate hain, to ye load ko
anycast routing, ensuring traffic optimization and automatic sambhal leta hai aur check karta hai ke
automated health-checking of destination servers. peeche server thik chal raha hai ya nahi.
2. Amazon CloudFront (Content Delivery Network - 2. Amazon CloudFront (CDN):
CDN): • Iska kaam website ki speed badhana hai. Ye
• A global web service that accelerates the delivery website ka data (images, text) apne paas
of static and dynamic web content (like HTML, temporary save (Cache) kar leta hai.
images, scripts) by caching it at edge locations • Door wale users (jaise Gwadar/Skardu) ko main
worldwide. server tak nahi jana parta; unhein unke sab se
• Drastically reduces user latency by serving static kareeb wali Edge Location se data foran mil jata
copies from nodes nearest to the user, preventing hai.
heavy traffic from overwhelming origin backend
databases.

Real-World Analogy (Asaan Misal)

English: Route 53 is the directory office that tells you where a shop is located. CloudFront is like a local branch
of a large store in your area so you don't have to travel to the central capital warehouse.
Roman English: Route 53 aik rasta batane wala board hai jo apko website ka address deta hai. CloudFront
apke sheher me bani aik choti branch hai taaki apko main factory tak safar na karna paray.

CS445: Distributed & Parallel Computing - Official Exam Notes Page 2 of 6

 Topic 2: Parallel Computing & CUDA Platform

PROFESSIONAL ENGLISH EASY ROMAN ENGLISH

1. Hardware Core Architecture (CPU vs GPU): 1. CPU vs GPU Farq:

• CPU: Comprises a few complex cores optimized for • CPU: Isme thode se par bohot zyada chalak cores
sequential tracking, deep logic, and executing hote hain. Ye muskhil logic aur ek ke baad ek
heavy single-threaded algorithms very fast. chalne wale kaamo ko bohot tez chalta hai.
• GPU: Built with thousands of smaller, simpler cores • GPU: Isme hazaron chote aur simple cores hote
running under SIMD (Single Instruction Multiple hain. Ye ek hi jaisa asaan kaam lakhon data points
Data), engineered to process repetitive parallel par aik hi waqt me (Parallelly) chalane ke liye bana
arrays at massive scale. hai.
2. Proprietary CUDA vs Open Ecosystems: 2. CUDA Limitations aur Dusre Options:
• CUDA: NVIDIA’s closed API. Code compiled in • CUDA: Ye sirf NVIDIA company ke graphics cards
CUDA cannot run on non-NVIDIA silicon, creating par chalta hai. Kisi AMD ya Intel card par iska code
strong vendor-lock frameworks. bilkul nahi chalega.
• Alternatives (OpenCL / AMD ROCm): Open- • Alternatives: OpenCL aur AMD ROCm open
source cross-platform runtime environments that platforms hain. Inka code AMD aur NVIDIA dono
allow massive data parallelism to execute cards par bina kisi masle ke chal jata hai.
flawlessly on AMD, Intel, and NVIDIA devices.

Real-World Analogy (Asaan Misal)

English: CPU is a group of 8 brilliant scientists solving a complex formula. GPU is a hall of 2,000 students doing
basic multiplication simultaneously.
Roman English: CPU 8 bade scientists ki tarah hai jo mushkil formula solve karte hain. GPU 2,000 bacho ki
tarah hai jo asaan jama-tafreeq (plus-minus) aik sath bohot jaldi kar lete hain.

CS445: Distributed & Parallel Computing - Official Exam Notes Page 3 of 6

 Topic 3: CAP Theorem in Distributed Databases

PROFESSIONAL ENGLISH EASY ROMAN ENGLISH

The CAP Core Trade-off Rules: CAP Theorem Ke Rules:

States that a distributed data system can guarantee at Distributed system me jab data centers ka aapas ka
most two out of three attributes when a network failure rabta toot jaye (Network Partition), to aap aik waqt me
occurs: sirf koi do cheezein bacha sakte hain:

• Consistency (C): Every read request receives the • Consistency (C): Poore network me sab ko bilkul
absolute latest data write, ensuring identical views naya aur 100% sahi data milega, kahin bhi galti
across all nodes. nahi hogi.
• Availability (A): Every non-failing node returns a • Availability (A): System hamesha chalega aur
response instantly without any errors, irrespective bina kisi error ke har user ko foran wapas jawab
of state updates. dega.
• Partition Tolerance (P): The system continues • Partition Tolerance (P): Agar machines ke beech
operating despite arbitrary connection drops or ka cable toot jaye, tab bhi system zinda rahega.
message cuts between node servers.
Bank vs Social Media Faisla: Kyunki network kabhi
Strict CP vs AP Choice: Real networks always bhi kharab ho sakta hai, isliye P zaroori hai. Banks
experience failure, meaning P is non-negotiable. CP model chunte hain (paise bachane ke liye system
Systems must choose to either lock down data access band kar denge par galat data nahi dikhaenge). Social
to preserve absolute safety (CP - Used by Banks) or Media apps AP model chunti hain (system chalta
serve stale/delayed data to maintain maximum system rahega, bhale hi like thoda late update ho).
uptime (AP - Used by Social Networks).

CS445: Distributed & Parallel Computing - Official Exam Notes Page 4 of 6

 Topic 4: Big Data Frameworks (Hadoop vs Apache Spark)

Hadoop MapReduce Framework Apache Spark Engine Framework

Feature
(Professional + Roman) (Professional + Roman)

Processing Disk-Based Architecture: Writes intermediate In-Memory Processing: Caches all analytical
Core calculation data to mechanical hardware drives. loops inside high-speed volatile RAM nodes.
Roman: Ye apni calculations baar baar physical Roman: Ye saara kaam computer ki temporary
hard disk par write karta hai jis se speed slow ho RAM me karta hai, jo hard disk se 100 guna zyada
jati hai. fast hoti hai.

Speed Ratio Standard sequential speed; best suited for non- Up to 100x faster execution due to minimized input/
urgent overnight data operations. output disk block cycles.
Roman: Normal processing speed hai. Bara data Roman: 100x tak fast processing hai kyunki isme
jo raat ko chalana ho uske liye thik hai. bar bar disk par write karne ka time zaya nahi hota.

Fault HDFS Data Replication: Splitting raw data into RDD (Resilient Distributed Datasets): Lineage
Tolerance blocks and copying them 3x across commodity graphs trace data source to recreate records if
nodes. nodes break.
Roman: HDFS 3x Replication use karta hai. Data Roman: Spark RDD use karta hai. Isko poora rasta
ke 3 hissay karke alag saste servers par copy kar yaad hota hai, agar data center ur jaye to ye dobara
deta hai. bana leta hai.

Topic 5: Filesystem in UserSpace (FUSE)

PROFESSIONAL ENGLISH EASY ROMAN ENGLISH

Core FUSE Mechanism: FUSE Kya Hai Aur Kaise Kaam Karta Hai:
• Traditional operating systems require filesystem • Aam taur par computer ka file chalane wala
drivers to exist inside the highly sensitive Kernel program system ke bilkul andar Kernel Space
Space. Any coding bug or software crash here (security vault) me hota hai. Agar wahan thodi si bhi
completely freezes the entire hardware system galti ho jaye, to poora computer freeze ya blue
(Kernel Panic). screen ho jata hai.
• FUSE Solution: Shifts filesystem logic out into the • FUSE Ka Solution: Ye code ko baahar User
secure User Space. Standard users can create Space me chalane ki ijazat deta hai. Ab developers
customized file structures (like virtual drives) safely. apni marzi ka naya custom file system (jaise
• A lightweight virtual bridge driver stays in the kernel Google Drive folder) asani se bana sakte hain.
to handle system calls, passing them securely • Agar is custom code me koi galti ho bhi jaye, to sirf
down to user program files. wo app band hogi, poora computer crash nahi
hoga.

CS445: Distributed & Parallel Computing - Official Exam Notes Page 5 of 6

 Topic 6: Parallel Scheduling Models

PROFESSIONAL ENGLISH EASY ROMAN ENGLISH

1. Static Scheduling Model: Tasks Ko Processors Par Schedule Karna:

• Tasks are mapped directly onto processors before • Static Scheduling: Program chalne se pehle hi
program initialization (Compile Time). (Compile Time par) faisla ho jata hai ke kaun sa
• Pros/Cons: Zero scheduling overhead during processor kaun sa kaam karega. Iska faida ye hai
execution. However, it results in critical load ke runtime par time zaya nahi hota, par agar koi
imbalances if some execution blocks take longer kaam lamba ho jaye to baki processors velle (idle)
than expected. baithe rehte hain.

2. Dynamic Scheduling Model: • Dynamic Scheduling: Program chalte waqt

(Runtime par) jaise jaise processors free hote hain,
• Tasks are allocated into a shared queue and fed to
unhein kaam diya jata hai. Iska faida ye hai ke koi
available processing units on the fly at Runtime.
free nahi baithta (good load balancing), par baar
• Pros/Cons: Maximizes resource load balancing baar list check karne me thoda system overhead
automatically. However, constant queue checking badh jata hai.
adds processing time overhead.

CS445: Distributed & Parallel Computing - Official Exam Notes Page 6 of 6