Foundational theory: what cloud computing is, the history of Amazon/AWS, what makes a true Cloud Service Provider, the CSP market landscape, compute abstraction levels (dedicated → VMs → containers → functions), the SaaS/PaaS/IaaS model, and cloud deployment models.

What is cloud computing

  • Definition: the practice of using a network of remote servers hosted on the internet to store, manage, and process data, rather than a local server or personal computer 8:54
  • On-premise: YOU own the servers, hire the IT people, pay/rent the real estate, take all the risks 8:54
  • Cloud provider: someone else owns servers, hires IT, pays real estate; you are only responsible for configuring cloud services and your code 9:18

Evolution of cloud hosting

  • Dedicated server (~1995): one physical machine, one business, one app — expensive (hardware, real estate, network, staff) but guaranteed high security; the model still exists for specialized use cases 9:38
  • Virtual Private Server (VPS): one physical machine subdivided into sub-machines via virtualization — better utilization and isolation, but you still bought a dedicated machine 10:09
  • Shared hosting (mid-2000s, GoDaddy/HostGator): one machine shared by hundreds of businesses via folders with permissions — very cheap, relied on tenants underutilizing resources, but limited functionality and poor isolation (one noisy tenant could hang every site on the server) 10:27 11:01
  • Cloud hosting: multiple physical machines acting as one system (distributed computing), abstracted into cloud services — flexible, scalable, secure (virtual isolation), very low cost (thousands of businesses share it), highly configurable; cloud still includes all prior hosting types as options 11:32 12:00

Amazon & AWS history

  • Amazon: American multinational tech corporation, headquartered in Seattle, founded 1994 by Jeff Bezos as an online bookstore 12:35
  • Expanded into cloud (AWS), streaming (Prime Video/Music), Twitch.tv, Whole Foods, AI, low-orbit satellites; current CEO is Andy Jassy (previously CEO of AWS) 13:17 13:41
  • AWS became a cloud service provider in 2006 and has led ever since; remember the term CSP (Cloud Service Provider) 14:34
  • First AWS service: SQS (Simple Queue Service) in 2004, then S3 in March 2006, then EC2 in August 2006 15:02
  • Amazon migrated all its own infrastructure to AWS by 2010 (later moving from Oracle to DynamoDB) 15:59
  • AWS launched its certification program in 2013, standardizing cloud certifications 16:30

What makes a Cloud Service Provider (CSP)

A company qualifies as a CSP when it has 17:10 17:34 18:06:

  • Multiple cloud services (tens to hundreds)
  • Services that can be chained together into cloud architectures
  • A single unified API (in AWS: the AWS API, accessed via CLI, SDK, Management Console)
  • Metered billing based on usage (per second/hour, vCPUs, memory, storage)
  • Rich monitoring built in — every API action tracked (AWS: CloudTrail)
  • An IaaS offering (networking, compute, storage, databases)
  • Automation via Infrastructure as Code

If a company offers multiple cloud services under one UI but doesn't meet most of these, it's a cloud platform (e.g., Twilio, HashiCorp, Databricks) — while AWS, Azure, GCP are cloud service providers 18:38

The CSP landscape (four tiers)

  • Top tier: AWS, Microsoft Azure, Google Cloud Platform, Alibaba Cloud — AWS/Azure/GCP are "the big three" in North America and Europe; Alibaba is huge in Asia/China 19:27; early to market with strong synergies between services 19:55
  • Mid tier: backed by well-known tech companies that didn't reach top tier — IBM (now expensive enterprise managed infrastructure) 20:27, Oracle (plays on being cheapest; even Azure signs contracts to use Oracle Cloud infrastructure) 20:49, plus Huawei Cloud and Tencent Cloud in Asia 21:27
  • Light tier: former VPS specialists offering core infrastructure — Vultr, DigitalOcean, Akamai Connected Cloud (formerly Linode) 21:51; lighter offerings (no event-driven metered billing depth) but simpler — a great cloud introduction for companies that find the top tier too complex 22:22
  • Private tier: software you deploy in your own data center for cloud-like functionality — OpenStack and Apache CloudStack (open source), VMware vSphere (widely used for managing VMs) 22:52 23:23

Gartner Magic Quadrant

  • Gartner's Magic Quadrant: market research reports by an IT consulting firm using proprietary qualitative analysis to show market direction, maturity, and participants 23:47
  • Four quadrants: Leaders, Challengers, Niche Players, Visionaries; AWS sits at the top of Leaders (May 2025 report), though Microsoft and Google have closed the gap — largely because of AI 24:32 27:27
  • Oracle keeps climbing, likely due to government contracts 26:40
  • Take it with a grain of salt: based on marketing-team questionnaires, self-certified metrics, and testimonials, with commercial incentives involved 28:11

The four core service types

  • The four most common cloud service types (the "four core" of IaaS): Compute, Networking, Storage, Databases; AWS has 200+ cloud services 28:45
  • Terminology note: the industry says "cloud computing" to mean ALL of cloud (compute, networking, storage, databases), not just the compute subcategory 29:23
  • The four core in AWS: Compute = EC2 (VMs), Storage = EBS (virtual hard drives), Database = RDS (SQL databases), Networking & Content Delivery = VPC (private cloud network) 29:50
  • Beyond the four core, categories include analytics, application integration, AR/VR, cost management, blockchain, containers, developer tools, IoT, ML, media, migration, quantum, robotics, satellites, security/identity/compliance, and more 30:26
  • Two ways to explore services: the marketing site aws.amazon.com (Products → overview / features / pricing / getting started / documentation) and the Management Console when logged in 30:59 31:46

Evolution of computing (compute abstraction levels)

LevelKey ideaTrade-offs
DedicatedPhysical server wholly used by one customer (single tenant); bare metal, you control virtualizationMust guess capacity upfront, overpay for underutilization, hard to vertically scale or replace; security guarantee is only as good as your skills; used for HPC 32:25 33:37 34:00
Virtual machinesMachine within a machine via a hypervisor; multi-tenant, pay a fraction of the serverStill overpay for underutilized VM; limited by guest OS but easy to export/import images and scale; the most common/popular compute offering 34:32 35:25
ContainersVM running containers via a container layer (Docker daemon; Docker most popular)Maximize capacity, easily add/resize containers; share the underlying OS, more efficient than multiple VMs; apps run side by side without OS conflicts — but more work to maintain 35:25 35:50
FunctionsServerless compute: upload only code, pick runtime (Ruby/Python/Node), memory and max run timeOnly pay while code runs, only responsible for code and data; trade-off is cold starts (VM spin-up can slow requests) 36:17 36:59

Types of cloud computing (SaaS / PaaS / IaaS)

  • SaaS (Software as a Service) — for customers: product fully run and managed by the provider; term first coined by Salesforce; examples: Gmail, Office 365 37:22 37:49
  • PaaS (Platform as a Service) — for developers: focus on deployment/management of apps without provisioning or understanding hardware/OS; examples: Elastic Beanstalk, Heroku, Google App Engine 37:49 38:16
  • IaaS (Infrastructure as a Service) — for administrators: basic building blocks of cloud IT (networking, computers, data storage) without worrying about IT staff, data centers, hardware; examples: AWS, Microsoft Azure, Oracle Cloud 38:43 39:10

Cloud deployment models

  • Public cloud: everything (workloads, projects, code) built on the CSP — known as "cloud-native" or "cloud-first" (note: "cloud-native" also has an older container/open-source meaning) 39:10 39:47
  • Private cloud: everything built in the company's own data center (on-premise), e.g., running OpenStack to launch VMs and databases 40:18
  • Hybrid cloud: on-premise + a cloud provider together, bridged via VPN or direct connection 40:41
  • Cross-cloud / multicloud: using multiple cloud providers at once (often mislabeled hybrid); e.g., Azure Arc extends the control plane to deploy Kubernetes on Amazon EKS or GCP Kubernetes Engine; GCP's equivalent is Anthos; AWS has traditionally not been cross-cloud friendly 41:12 41:43

Who uses which deployment model

  • Cloud: startups, SaaS companies, new projects — e.g., Basecamp, Dropbox, Squarespace 42:23 42:54
  • Hybrid: organizations with existing data centers that can't fully migrate due to effort or security/compliance — banks, fintech, investment management (e.g., CIBC, Deloitte, CPP Investment Board) 42:54
  • On-premise: strict regulatory compliance or sheer size/politics — public sector/government, hospitals, heavily regulated insurance (e.g., AIG, Government of Canada); though even these use public-sector cloud offerings — there's rarely a good reason to be fully on-premise anymore 43:21 43:48