Unit_1_Computing_Paradigm

Loading views...

1.1 Introduction to Cloud Computing

1.1.1 The Historic Vision (1969)

In 1969, Leonard Kleinrock — one of the chief scientists of ARPANET (the predecessor to the Internet) — made a remarkable prediction:

"As of now, computer networks are still in their infancy, but as they grow up and become sophisticated, we will probably see the spread of 'computer utilities' which, like present electric and telephone utilities, will service individual homes and offices across the country."

He predicted that computers would become a utility — just like electricity or water — more than 30 years before Cloud Computing became a reality. This is the foundational vision of cloud computing.

The Modern Equivalent:

"I don't care where my servers are, who manages them, where my documents are stored, or where my applications are hosted. I just want them always available and access them from any device connected through the Internet. And I am willing to pay for this service for as long as I need it."

1.1.2 What is Cloud Computing?

Cloud Computing is the paradigm of delivering IT services on demand over the Internet, on a pay-per-use basis, just like a utility service.

Three Key Definitions You Must Know:

① Armbrust et al.:

"Cloud computing refers to both the applications delivered as services over the Internet, and the hardware and system software in the datacenters that provide those services."

Covers the entire stack: from bare hardware up to end-user applications.
Introduces XaaS (Everything as a Service).

② NIST (National Institute of Standards and Technology) — The Official 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."

③ Buyya et al. — A Technical Definition:

"A Cloud is a type of parallel and distributed system consisting of a collection of interconnected and virtualized computers that are dynamically provisioned and presented as one or more unified computing resources based on service-level agreements established through negotiation between the service provider and consumers."

1.1.3 Reese's 3 Criteria: "Is it truly Cloud?"

According to Reese, a service qualifies as Cloud Computing only if it meets all three criteria:

#	Criterion	What it means	Example
1	Accessible via Web Browser or Web API	No proprietary client software needed	Access Gmail from any browser
2	Zero Capital Expenditure (CapEx) to start	No upfront hardware purchase required	Sign up for AWS with just a credit card
3	Pay only for what you use	Costs are tied strictly to usage, not flat rates	AWS EC2 billed per second of VM uptime

1.1.4 Key Concepts in Cloud Computing

  +----------------------------------------------------------+
  |                    CLOUD COMPUTING                       |
  |                                                          |
  |  [No upfront costs]  [Pay-per-use]  [Elastic Scaling]    |
  |  [Access anywhere]   [Virtualized]  [Managed by vendor]  |
  |                                                          |
  |  Technologies inside the "Cloud":                        |
  |    - Virtualization     - Web 2.0                        |
  |    - Service Orientation- Utility Computing              |
  |    - Distributed Systems                                 |
  +----------------------------------------------------------+
         |           |           |
       IaaS         PaaS        SaaS
  (Infrastructure)(Platform) (Software)

1.2 The Cloud Computing Reference Model (SPI Model)

Cloud computing services are organized into three layers — IaaS, PaaS, and SaaS. Together they form the SPI Model.

  +----------------------------------------------+
  |              SaaS                            |
  |  (End-user Applications)                     |
  |  Google Docs, Salesforce, Facebook, Flickr   |
  +----------------------------------------------+
  |              PaaS                            |
  |  (Runtime + Development Platform)            |
  |  Google AppEngine, Windows Azure, Aneka      |
  +----------------------------------------------+
  |              IaaS                            |
  |  (Virtual Hardware + Storage + Networking)   |
  |  Amazon EC2, S3, RightScale, VMware vCloud   |
  +----------------------------------------------+
  |         Physical Datacenter                  |
  |     (Servers, Storage, Network Fabric)       |
  +----------------------------------------------+

1.2.1 Infrastructure-as-a-Service (IaaS)

What is it? You rent raw, virtualized computing infrastructure — virtual machines, storage, and networks. You control everything from the OS upward; the provider manages the physical hardware.

What you get:

Virtual Compute: VM instances (e.g., 2 vCPUs, 8 GB RAM). You choose OS, install software.
Virtual Storage:

- Raw Disk: A virtual hard drive attached to your VM (like a real HDD).

- Object Store: Store files/blobs as objects (e.g., Amazon S3). No file system — just key-value (name → file).

Virtual Networking: Firewall rules, private networks (VPCs), load balancers, DNS.

Pricing Model: Usually $/hour based on VM size (CPU, RAM, storage).

Who uses IaaS?

Developers building scalable web sites.
Companies doing batch data processing.
Anyone who needs a server but doesn't want to buy one.

Examples: Amazon EC2, Amazon S3, RightScale, VMware vCloud, Microsoft Azure VMs.

1.2.2 Platform-as-a-Service (PaaS)

What is it? A managed development and hosting platform. You write code; the platform handles deployment, scaling, fault tolerance, and infrastructure. You never touch a server.

What you get:

A runtime environment for your app (Python, Java, Node.js, etc.)
Auto-scaling (more traffic? More instances automatically created)
Built-in services: databases, caches, queues, email sending
Development SDK + testing tools

What the provider manages:

Servers, OS, middleware, runtime, scaling — everything below your code.

Who uses PaaS?

Developers who want to focus purely on application logic.
Startups that can't afford a DevOps team.

Examples: Google App Engine (Python, Java, Go), Windows Azure (roles model), Force.com (Salesforce apps), Heroku, Aneka.

Trade-off: Higher abstraction = less control. You are constrained to the provider's supported languages and frameworks. Vendor lock-in is a concern.

1.2.3 Software-as-a-Service (SaaS)

What is it? Complete, ready-to-use applications delivered over a browser. No installation, no maintenance.

What you get:

Full application accessible from any browser.
Multi-tenancy: one application instance serves thousands of users, each isolated from the others.
Auto-updates: the provider rolls out new features without you doing anything.

Who uses SaaS?

End users (documents, email, photos).
Businesses (CRM, accounting, HR).

Examples: Google Docs (office), Salesforce (CRM), Facebook (social), Flickr (photos), Gmail (email).

1.2.4 Comparison: IaaS vs. PaaS vs. SaaS

Feature	IaaS	PaaS	SaaS
You manage	OS, App, Data	App, Data	Nothing
Provider manages	Hardware, Network	Hardware, OS, Runtime	Everything
Control level	Highest	Medium	Lowest
Ease of use	Requires expertise	Moderate	Very easy
Target user	System/infra engineers	App developers	End users
Example	AWS EC2	Google AppEngine	Gmail
Pricing	$/hour per VM	$/request or resources used	$/month subscription

1.3 NIST 5 Essential Characteristics

The NIST definition defines 5 essential characteristics that every true cloud service must have:

#	Characteristic	Explanation	Example
1	On-demand Self-Service	Users provision resources themselves without human interaction with the provider	Launch an EC2 VM from the AWS console at 2 AM without calling Amazon
2	Broad Network Access	Available over the Internet via standard mechanisms (HTTP, REST)	Access Google Docs from phone, tablet, laptop
3	Resource Pooling	Provider serves many customers from a shared pool (multi-tenancy). Customer doesn't know which physical server they're on	1000 customers share 50 physical servers; each VM is isolated
4	Rapid Elasticity	Resources can be scaled up or down quickly, even automatically	Auto-scaling group adds 10 VMs when CPU > 80%, removes them when load drops
5	Measured Service	Usage is monitored, controlled, and reported transparently. Billing is based on actual consumption	AWS bills by the second; you see exact resource usage in your dashboard

1.4 Cloud Computing Deployment Models

  +------------------+    +------------------+    +------------------+
  |   PUBLIC CLOUD   |    |  PRIVATE CLOUD   |    |   HYBRID CLOUD   |
  |                  |    |                  |    |                  |
  | 3rd party owned  |    | Org-owned infra  |    | Mix of public +  |
  | Anyone can use   |    | Internal use only|    | private          |
  | Multi-tenant     |    | High control     |    | Cloudbursting    |
  |                  |    |                  |    |                  |
  | AWS, GCP, Azure  |    | OpenStack, vCloud|    | AWS + On-premise |
  +------------------+    +------------------+    +------------------+

Model	Owned By	Users	Data Location	Use Case
Public	3rd party provider	General public	Provider's datacenters	Startups, general workloads
Private	The organization	Internal only	On-premise	Banks, government, hospitals
Hybrid	Both	Internal + public overflow	Both	Peak-load handling (cloudbursting)
Community	Shared by multiple orgs	Member orgs	Shared facility	Healthcare, government agencies

Cloudbursting: When private cloud hits capacity during peak demand, workloads automatically "burst" to a public cloud.

  Normal Load:   [Private Cloud] --> handles all traffic
  Peak Load:     [Private Cloud] --> overflow --> [Public Cloud] (burst)

1.5 Benefits of Cloud Computing

Economic Benefits

CapEx to OpEx Transformation:

  BEFORE CLOUD (CapEx model):
  Year 1: Buy $500,000 worth of servers (upfront capital expenditure)
          Server is worth $400,000 by Year 2 (depreciation!)
          You still pay maintenance + admin even if server sits idle

  AFTER CLOUD (OpEx model):
  Year 1: Pay $50,000 for what you actually used
          No depreciation. No idle cost. No upfront investment.
          Shut down resources when not needed → $0 cost.

Benefit	Explanation
No upfront cost	Start a business with just a credit card
Pay-per-use	Pay only for what you actually consume
No depreciation	No hardware to depreciate or replace
Lower admin costs	Provider handles maintenance, updates, hardware
Economies of scale	Provider buys hardware at bulk discount; passes savings to users

Technical Benefits

Benefit	Explanation
Elasticity	Scale up instantly for a traffic spike; scale down immediately after
High Availability	Provider's SLA guarantees 99.9%+ uptime
Multi-tenancy efficiency	Sharing infrastructure across thousands of users = better utilization
Agility	Deploy a new server in minutes, not weeks
Green IT	Consolidation reduces total energy consumption across the industry

1.6 Real-World Case Studies

Case A: New York Times (Large Enterprise)

Problem: Convert 11 million newspaper articles (1851–1922) from TIFF images to PDFs. Required massive compute power for a short period.
Solution: Rented Amazon EC2 and S3 compute + storage.
Result: Task completed in 36 hours, paying only for the duration. Cost was tiny compared to buying dedicated servers for a one-time job.
Lesson: Cloud is ideal for bursty, short-duration, high-compute workloads.

Case B: Animoto (Startup)

Problem: Animoto creates personalized videos from photos and music. Processing is storage and CPU intensive. Load is unpredictable.
Solution: Entire backend built on Amazon Web Services — no owned servers.
Result: When a Facebook campaign went viral, Animoto scaled from 50 to 3,500 servers in 3 days. Without cloud, this would be impossible.
Lesson: Cloud enables startups to handle massive, unpredictable growth without upfront investment.

Case C: Little Fluffy Toys (Developer)

Problem: Company built a widget for finding nearby bicycle rental services. Needed a scalable backend.
Solution: Hosted on Google App Engine.
Result: App was on the market in just one week.
Lesson: PaaS platforms dramatically reduce time-to-market.

Case D: Apple iCloud (End User)

Problem: Users want their photos/documents available on every device — iPhone, laptop, iPad — without manually syncing.
Solution: Files stored in the cloud, automatically synced to all devices.
Result: Take a photo on your iPhone → it's instantly on your MacBook and iPad.
Lesson: Cloud enables seamless, transparent device-agnostic access to personal data.

1.7 Challenges of Cloud Computing

Challenge	Explanation
Security	Data must be stored on the provider's servers. If a malicious provider captures VM memory, even encrypted data can be read (it's decrypted in RAM for processing).
Privacy	Who has legal access to your data? Jurisdiction matters.
Legal/Regulatory	American Patriot Act gives US government access to data. EU privacy laws are stricter. If a US company stores data in Europe, conflicting laws create a legal grey zone.
Vendor Lock-in	Proprietary APIs and formats make it difficult to migrate from one provider to another.
Performance	Network latency is added to every request. Shared infrastructure can cause "noisy neighbor" problems.
Availability Dependency	If the cloud provider goes down (AWS outage), your application goes down too.

1.8 Historical Development of Cloud Computing

1.8.1 Timeline of Distributed Computing

1951 ──> UNIVAC I: First commercial mainframe
1960 ──> Cray's first supercomputer
1966 ──> Flynn's Taxonomy (SISD, SIMD, MISD, MIMD) for parallel computing
1969 ──> ARPANET: The seed of the Internet (Kleinrock's "computer utility" quote)
1970 ──> DARPA's TCP/IP protocol development begins
1975 ──> Xerox PARC invents Ethernet
1984 ──> DEC VMScluster + IEEE 802.3 (Ethernet & LAN standardized)
1989 ──> TCP/IP standardized (IETF RFC 1122)
1990 ──> Lee-Cailliau: WWW, HTTP, HTML — the web is born
1997 ──> IEEE 802.11 (Wi-Fi standard)
1999 ──> Grid Computing + Amazon begins (later AWS)
2004 ──> Web 2.0 era begins
2005 ──> Amazon Web Services launched (EC2, S3)
2007 ──> Manjrasoft Aneka platform
2008 ──> Google App Engine launched
2010 ──> Microsoft Azure becomes generally available

1.8.2 Mainframe Computing (1950s)

What: Large, powerful centralized computers. Multiple processors in one cabinet. Presented as a single entity to users.

Key Features:

Specialized for massive I/O operations and bulk data processing.
Always-on: Failed components could be replaced without shutting down the system (hot-swap).
No distributed computing — everything is centralized.

Applications (still in use today): Online banking (SWIFT), airline booking (SABRE), government services.

Relevance to Cloud: Clouds inherit the "always-on" and "single logical entity" characteristics of mainframes.

1.8.3 Cluster Computing (1980s)

What: Groups of commodity PCs/servers connected by high-bandwidth networks and managed as a single system.

  [ Node 1 ] --+
  [ Node 2 ] --+-- [ High-Speed Network ] --> [ Job Scheduler ] --> Users
  [ Node 3 ] --+
  [ Node N ] --+
  
  Each node is a cheap commodity machine. Together, they = a supercomputer.

Driver: Cheap commodity hardware became powerful enough to tackle HPC workloads at a fraction of mainframe cost.

Key Tools:

Condor: Job scheduling for clusters.
PVM (Parallel Virtual Machine): Run parallel jobs across heterogeneous machines.
MPI (Message Passing Interface): Standard API for nodes to communicate in parallel programs (still dominant today).

Relevance to Cloud: Cloud provider datacenters ARE massive clusters. Hadoop runs on commodity cluster nodes.

1.8.4 Grid Computing (1990s)

What: Aggregation of geographically dispersed clusters from multiple organizations, connected via the Internet, forming a planet-scale compute resource.

  University A (100 nodes)
          |
          +------ INTERNET ------ National Lab (500 nodes) -----+
          |                                                       |
  University B (200 nodes) -------------------------------- Research Institute (300 nodes)
  
  All combined = 1100 nodes available on demand via a Grid middleware

Key Differences from a Cluster:

Feature	Cluster	Grid
Location	Same room/building	Geographically distributed
Ownership	One organization	Multiple organizations
Homogeneity	Usually same hardware	Heterogeneous
Scale	Hundreds of nodes	Thousands to millions
Admin domain	One	Many (Virtual Organizations)

Why Grids emerged:

Clusters were now common but underutilized.
New problems required compute beyond any single cluster.
Internet made long-distance high-bandwidth connectivity possible.

Examples: CERN LHC Grid (particle physics), BIRN (biomedical research), IVOA (astronomy).

Relevance to Cloud: Grids introduced the utility computing vision — pay-per-use of shared resources. Cloud refines this with easier access and commercial pricing.

1.8.5 Cloud vs. Grid vs. Cluster: Key Differences

Feature	Cluster	Grid	Cloud
Access	Usually private	Via middleware (complex)	Via Web browser / REST API
Pricing	Free (internal)	Free / allocation-based	Pay-per-use (commercial)
Setup complexity	High	Very High	Near zero
Elasticity	None (fixed nodes)	Limited	Near-infinite
Virtualization	Rarely used	Sometimes	Core technology
SLA guarantee	Informal	Best-effort	Commercial SLA

1.9 Enabling Technologies

1.9.1 Virtualization

Allows one physical server to run many isolated VMs.
Enables IaaS (rent a VM, not a physical server).
Enables resource pooling and multi-tenancy.

1.9.2 Web 2.0

Transformed the web from static pages (Web 1.0) to interactive applications.
Technologies: AJAX, REST APIs, JSON, HTML5.
Made it viable to run full applications (Google Docs) in a browser.

1.9.3 Service-Oriented Computing (SOC)

Core model: services as building blocks of applications.
Services are: loosely coupled, reusable, platform-independent, location-transparent.
Web Services: Use HTTP + SOAP + WSDL to expose software as accessible services.
REST (Representational State Transfer): Simpler alternative to SOAP. Uses HTTP verbs (GET, POST, PUT, DELETE).
SOC introduces QoS and SLA concepts — important for cloud billing.
SOC introduces SaaS — delivering complete software functionality as a service.

1.9.4 Utility Computing

Vision: package compute resources (CPU, storage, apps) on a pay-per-use basis.
First articulated by John McCarthy (1961): "Computing may someday be organized as a public utility just as the telephone system..."
Grids implemented this at academic scale. Cloud implements it commercially.

1.10 Major Cloud Platforms (Reference Implementations)

1.10.1 Amazon Web Services (AWS)

Type: IaaS (and more)
Core Services:

- EC2 (Elastic Compute Cloud): Virtual machines on demand. Choose CPU, RAM, disk, OS.

- S3 (Simple Storage Service): Object storage. Store files as objects in "buckets".

Additional Services: RDS (database), VPC (networking), Route53 (DNS), CloudFront (CDN), ElasticCache, SQS (queuing), etc.
Significance: First major commercial cloud. Launched 2005. Still market leader.

1.10.2 Google App Engine (GAE)

Type: PaaS
What it does: Host Python, Java, or Go web applications. Auto-scales automatically.
Built-in services: Scalable Datastore (NoSQL), Memcache, Task Queues, Cron Jobs.
Developer workflow: Build locally using SDK → Deploy with one command → Google handles all scaling.

1.10.3 Microsoft Azure

Type: PaaS + IaaS
Roles model:

- Web Role: Hosts a web application (IIS-based).

- Worker Role: Background processing (e.g., video encoding, batch jobs).

- VM Role: Full customizable virtual machine.

Additional Services: Azure SQL, Azure Storage (Blobs), Azure Service Bus, CDN.

1.10.4 Apache Hadoop

Type: Open-source data processing platform (used for PaaS-level big data)
Based on: Google's MapReduce paper (2004) and Google File System (GFS).
Two core components:

- HDFS (Hadoop Distributed File System): Stores data across many nodes with replication.

- Hadoop MapReduce: Processes data in parallel using Map + Reduce functions.

Sponsor: Yahoo! runs the world's largest Hadoop cluster.

1.10.5 Force.com / Salesforce.com

Type: SaaS + PaaS
What it does: A complete platform for building social enterprise applications.
Salesforce.com is built on Force.com — the #1 SaaS CRM in the world.
Marketplace: AppExchange — ready-made components you can plug into your Salesforce app.

1.11 Comparison Summary

Platform	Model	Primary Use	Language	Key Feature
AWS EC2	IaaS	Virtual servers	Any	Most flexible, biggest ecosystem
Google App Engine	PaaS	Web apps	Python, Java, Go	Auto-scales to zero and to millions
Microsoft Azure	PaaS+IaaS	Enterprise apps	.NET, Java, etc.	Deep Windows/Microsoft integration
Hadoop	Data Platform	Big Data processing	Java	MapReduce on commodity hardware
Force.com	SaaS+PaaS	CRM / Enterprise apps	Apex (proprietary)	Pre-built enterprise components

Navigation

Next Unit: Parallel and Distributed Computing

Unit_1_Computing_Paradigm

Unit 1: Overview of Computing Paradigm

1.1 Introduction to Cloud Computing

1.1.1 The Historic Vision (1969)

1.1.2 What is Cloud Computing?

1.1.3 Reese's 3 Criteria: "Is it truly Cloud?"

1.1.4 Key Concepts in Cloud Computing

1.2 The Cloud Computing Reference Model (SPI Model)

1.2.1 Infrastructure-as-a-Service (IaaS)

1.2.2 Platform-as-a-Service (PaaS)

1.2.3 Software-as-a-Service (SaaS)

1.2.4 Comparison: IaaS vs. PaaS vs. SaaS

1.3 NIST 5 Essential Characteristics

1.4 Cloud Computing Deployment Models

1.5 Benefits of Cloud Computing

Economic Benefits

Technical Benefits

1.6 Real-World Case Studies

Case A: New York Times (Large Enterprise)

Case B: Animoto (Startup)

Case C: Little Fluffy Toys (Developer)

Case D: Apple iCloud (End User)

1.7 Challenges of Cloud Computing

1.8 Historical Development of Cloud Computing

1.8.1 Timeline of Distributed Computing

1.8.2 Mainframe Computing (1950s)

1.8.3 Cluster Computing (1980s)

1.8.4 Grid Computing (1990s)

1.8.5 Cloud vs. Grid vs. Cluster: Key Differences

1.9 Enabling Technologies

1.9.1 Virtualization

1.9.2 Web 2.0

1.9.3 Service-Oriented Computing (SOC)

1.9.4 Utility Computing

1.10 Major Cloud Platforms (Reference Implementations)

1.10.1 Amazon Web Services (AWS)

1.10.2 Google App Engine (GAE)

1.10.3 Microsoft Azure

1.10.4 Apache Hadoop

1.10.5 Force.com / Salesforce.com

1.11 Comparison Summary

Navigation