📌 Lecture 5 — Virtualization: One Box, Many Worlds

📍 Slide 1 – 💥 The Day VMware Saved a Datacenter

🗓️ 2002 — Diane Greene's team at VMware demos ESX Server, the first production hypervisor for x86
🪦 Before ESX: a typical datacenter ran 5-10% CPU utilization because every app needed a dedicated box "to be safe"
📈 After ESX: enterprises consolidated 10:1, 20:1, 50:1 — entire physical rooms emptied of servers, costs dropped 70%
🚀 The pattern made AWS possible: by August 2006, EC2 launched, renting virtual machines by the hour
🎓 Lesson: Software pretending to be hardware unlocked the entire cloud era

🤔 Think: Before virtualization, "deploying QuickNotes" meant buying a server, racking it, and installing Linux. How would your weekend look?

📍 Slide 2 – 🎯 Learning Outcomes

#	🎓 Outcome
1	✅ Distinguish Type 1 vs Type 2 hypervisors with examples
2	✅ Explain what a VM image actually is (and isn't)
3	✅ Snapshot, restore, clone — the three superpowers a VM gives you
4	✅ Compare VM vs container (preview of Lecture 6)
5	✅ Spin a Linux VM in VirtualBox and SSH into it
6	✅ Use Vagrant to make VM setup reproducible (declarative, in Git)

📍 Slide 3 – 🗺️ Lecture Overview

graph LR
    A["💾 What is a VM?"] --> B["🏛️ Type 1 vs Type 2"]
    B --> C["🛠️ Lifecycle"]
    C --> D["📦 Images & Snapshots"]
    D --> E["📜 Vagrant"]
    E --> F["🐳 VM vs Container"]

📍 Slides 1-7 — VM fundamentals
📍 Slides 8-12 — Lifecycle, images, snapshots
📍 Slides 13-16 — Vagrant + the QuickNotes VM
📍 Slides 17-21 — VM vs container, real story, takeaways

📍 Slide 4 – 💾 What Is a Virtual Machine, Really?

A VM is a software emulation of a complete computer — its own CPU, RAM, disks, NIC — running on top of real hardware.

graph TB
    subgraph "Host"
        H1["🔧 Hardware (CPU, RAM, Disks, NIC)"] --> HV["🪞 Hypervisor"]
        HV --> V1["💻 VM 1<br/>Ubuntu + QuickNotes"]
        HV --> V2["💻 VM 2<br/>Windows + IIS"]
        HV --> V3["💻 VM 3<br/>Alpine + DB"]
    end

🪞 The hypervisor is the magic layer that fools each VM into thinking it has the whole machine
🆔 The VM has its own kernel — that's the key difference from a container (next lecture)
⏳ Hardware features make this fast: Intel VT-x (2005), AMD-V (2006), nested page tables (2008)

📍 Slide 5 – 🏛️ Type 1 vs Type 2 Hypervisors

	Type 1 (Bare-metal)	Type 2 (Hosted)
Where it sits	Directly on hardware	On top of a host OS
Examples	VMware ESXi, Microsoft Hyper-V, KVM (Linux kernel), Xen	VirtualBox, VMware Workstation, Parallels
Performance	Closer to bare metal	A few % slower (host OS overhead)
Use case	Datacenter / cloud	Developer laptops

graph LR
    subgraph "Type 1"
        HW1["Hardware"] --> HV1["Hypervisor (KVM)"]
        HV1 --> VM1["VM 1"]
        HV1 --> VM2["VM 2"]
    end
    subgraph "Type 2"
        HW2["Hardware"] --> OS2["Host OS"]
        OS2 --> HV2["VirtualBox"]
        HV2 --> VMa["VM"]
    end

🧪 In this lab you'll use VirtualBox (Type 2) — free, cross-platform, perfect for learning
🏛️ AWS EC2, GCP Compute, every "cloud VM" is KVM (Type 1) behind the scenes

📍 Slide 6 – 📜 A Short History of x86 Virtualization

🖥️ 1972 — IBM ships VM/370 on mainframes — virtualization was already old then
🤔 1999 — VMware founders prove you can virtualize x86 even though the architecture wasn't designed for it (binary translation)
🏎️ 2005-2006 — Intel VT-x and AMD-V add hardware support → 2-5× speedup
🐧 2007 — KVM merges into the Linux kernel; suddenly every Linux box is a hypervisor
☁️ 2006-2008 — AWS launches EC2, S3, RDS — the cloud era is here
🧪 2013-2015 — Docker and Kubernetes start to eat the VM's lunch (Lecture 6)

💡 Virtualization didn't replace bare metal — it added an option. Cloud + containers didn't replace VMs — they added more options. Pick the right tool, not the new tool.

📍 Slide 7 – 🛠️ VM Lifecycle: From Image to Running

stateDiagram-v2
    [*] --> Image: download / build
    Image --> Created: clone template
    Created --> Running: power on (boot)
    Running --> Paused: suspend
    Paused --> Running: resume
    Running --> Stopped: shutdown
    Running --> Snapshot: snapshot
    Snapshot --> Running: restore
    Stopped --> [*]

State	What it is	Disk footprint
Image	Cold file on disk (.vdi, .qcow2, .ova)	Base size
Created	VM exists in inventory but never booted	Same as image
Running	CPU + RAM allocated, OS executing	+ RAM size on disk if paged
Snapshot	Copy-on-write fork of the disk	Δ from base
Paused	RAM dumped to disk, VM frozen	+ RAM file

📍 Slide 8 – 📦 VM Images & Disk Formats

Format	Origin	Compression	Snapshots
`.vdi`	VirtualBox	✅	✅
`.qcow2`	QEMU/KVM	✅	✅ (copy-on-write)
`.vmdk`	VMware	✅	✅
`.vhd / .vhdx`	Microsoft	✅	✅
`.ova / .ovf`	Open Virtualization Format	✅ (tar of VM + metadata)	—

📁 An image is a file. You can copy it, version it, scan it
🆔 Cloud images (Ubuntu cloud-images, Amazon Linux) are pre-baked with cloud-init so they boot, accept your SSH key, install your packages, with no manual interaction
🧰 qemu-img convert -f qcow2 -O vdi src.qcow2 dst.vdi — convert between formats

📍 Slide 9 – 📸 Snapshots: The Time Machine

# VirtualBox CLI
$ VBoxManage snapshot quicknotes-vm take pre-deploy
$ VBoxManage snapshot quicknotes-vm restore pre-deploy
$ VBoxManage snapshot quicknotes-vm delete pre-deploy

⏪ A snapshot freezes the VM's disk + memory state at a moment in time
🪤 Copy-on-write — the snapshot doesn't duplicate the disk; only changed blocks are written
⚠️ Snapshots are not backups — they live on the same disk. A failing disk takes both
🚫 Don't run a VM on dozens of snapshots — every read traverses the chain. Snapshot, restore, delete

💡 In Lab 5, you'll snapshot a clean QuickNotes VM, deliberately break it, restore in 30 seconds.

📍 Slide 10 – 🔗 VM Networking: Three Modes

Mode	What it does	When to use
NAT	VM gets a private IP, shares host's IP for outbound	Default; just need internet from VM
Bridged	VM gets an IP on the same LAN as the host	Other machines on the LAN can reach the VM
Host-only	Isolated network between host ↔ VM (no internet)	Lab environments, secret testing

# VirtualBox: tell the VM to forward host port 8080 → VM port 8080
$ VBoxManage modifyvm quicknotes-vm \
    --natpf1 "quicknotes,tcp,127.0.0.1,8080,,8080"

🧪 Lab 5 uses NAT with port forwarding so curl localhost:8080/notes from your host reaches QuickNotes inside the VM

📍 Slide 11 – 📜 Vagrant: Make VMs Reproducible

Vagrant turns "click through 27 GUI screens" into a single text file in Git:

# Vagrantfile
Vagrant.configure("2") do |config|
  config.vm.box      = "ubuntu/jammy64"
  config.vm.hostname = "quicknotes-vm"
  config.vm.network "forwarded_port", guest: 8080, host: 18080

  config.vm.provider "virtualbox" do |vb|
    vb.memory = 1024
    vb.cpus   = 2
  end

  config.vm.provision "shell", inline: <<-SHELL
    apt-get update && apt-get install -y curl
  SHELL
end

vagrant up        # build + boot
vagrant ssh       # log in
vagrant snapshot save clean-quicknotes
vagrant destroy   # nuke it from orbit

📝 One file = one reproducible VM. Commit it. Diff it. Code-review it
🧪 Lab 5 ships you a Vagrantfile as plumbing; you'll customize provisioning

📍 Slide 12 – 🚀 cloud-init: The Other Half of Reproducibility

For cloud VMs, cloud-init is the universal first-boot configurator:

#cloud-config
hostname: quicknotes
users:
  - name: deploy
    ssh-authorized-keys: [ssh-ed25519 AAAA...]
    sudo: ALL=(ALL) NOPASSWD:ALL
packages:
  - curl
  - jq
runcmd:
  - systemctl enable --now quicknotes
write_files:
  - path: /etc/quicknotes/env
    content: |
      ADDR=:8080

🌐 Supported by AWS, GCP, Azure, OpenStack, DigitalOcean — every cloud
🧪 Lab 10 will use cloud-init to spin a Cloud Run-adjacent test VM
📚 The same YAML works in Vagrant too

📍 Slide 13 – 📈 Resource Sizing: How Much Is Enough?

Workload	vCPU	RAM	Disk
Idle Linux + sshd	1	256 MB	1 GB
QuickNotes (this course)	1	512 MB	2 GB
Postgres dev DB	2	2 GB	10 GB
Build/CI runner	4	8 GB	20 GB

📊 Over-provisioning wastes money in the cloud and host RAM on your laptop
📉 Under-provisioning triggers OOM kills, swap thrashing, mystery slowness
🧰 free -h, vmstat 1, iostat 1 — measure before you guess
💡 In Lab 8 (SRE) you'll learn to size based on observed load, not guessed load

📍 Slide 14 – 🐳 VM vs Container: The 30-Second Preview

	Virtual Machine	Container
Has its own kernel?	✅ Yes	❌ Shares host kernel
Boot time	30-90 s	1-3 s
Disk size	1-10 GB	5-200 MB
Strong isolation	✅ Hardware-level	⚠️ Namespace + cgroup level
Lifetime	Days to years	Seconds to days
Use it for	Multi-tenant clouds, full OS choice	Microservices, CI jobs, dev environments

graph LR
    subgraph "VM"
        HV["Hypervisor"] --> KG["Guest kernel 1"]
        HV --> KG2["Guest kernel 2"]
    end
    subgraph "Containers"
        K["Shared host kernel"]
        K --> C1["containerd 1"]
        K --> C2["containerd 2"]
        K --> C3["containerd 3"]
    end

🤝 In production, containers usually run inside VMs (cloud-provided isolation + container density)
🎯 Next lecture: containers in depth

📍 Slide 15 – 🧪 Lab 5 Preview: QuickNotes in a VM

🛠️ Task 1 (6 pts): vagrant up a Ubuntu 24.04 VM, install Go, build QuickNotes inside, expose :8080 to the host, hit it with curl from your laptop
📸 Task 2 (4 pts): Snapshot the clean VM, deliberately break the binary, restore the snapshot, measure recovery time
🎁 Bonus (2 pts): Compare boot time + RAM footprint of a Vagrant VM vs running QuickNotes in a Docker container (preview of Lab 6)
📜 Deliverable: submissions/lab5.md with vagrant up log, snapshot timestamps, and a one-paragraph reflection

📍 Slide 16 – ❌ Common VM Antipatterns

🔥 Antipattern	✅ Better
GUI-click your way through VM setup, then "remember" what you did	Vagrantfile + cloud-init = version-controlled
Run a VM on a chain of 20 snapshots	Use snapshots transiently; delete after restore
Allocate 16 vCPU on a 4-core host	Over-provisioning leads to thrashing — cap at host cores
Skip `apt-get update` in provisioning ("the image was new last month")	Build images with Packer; pin versions; refresh weekly
Treat VM as a pet ("this is the build server")	Treat VM as cattle — destroyable + rebuildable in one command

📍 Slide 17 – 📜 Real Story: Heartbleed and the Patch Storm

🗓️ April 7, 2014 — Heartbleed (CVE-2014-0160) disclosed: any TLS-using OpenSSL server can be remotely memory-dumped
🧨 ~17% of the public internet's HTTPS servers were vulnerable
🏃 At cloud scale, this meant rebuilding tens of thousands of VM images from base + new OpenSSL + redeploying
🛠️ Companies with proper image-pipelines (Packer + version-controlled cloud-init) shipped patches in hours. Companies that hand-built VMs took weeks
🎓 Lesson: If you can't rebuild a VM from a text file, you can't respond to a CVE at speed

💬 "Treat your servers like cattle, not pets." — Bill Baker (Microsoft, 2012) — and Heartbleed was the moment that became survival, not philosophy

📍 Slide 18 – 🧠 Key Takeaways

💾 A VM is software pretending to be hardware — its own kernel, its own networking, its own everything
🏛️ Type 1 = bare-metal hypervisor (datacenter), Type 2 = hosted (laptop) — same idea, different host
📸 Snapshots are time machines, not backups — same disk, same failure domain
📜 Vagrant + cloud-init make VMs declarative — commit the file, not the clicks
🐳 VMs and containers are complementary, not competitive — containers run inside VMs in production
🐄 Treat servers like cattle — when you can rebuild from text, you can survive Heartbleed-class events

📍 Slide 19 – 🚀 What's Next + 📚 Resources

📍 Next lecture: Containers — same OS, lighter isolation, faster everything
🧪 Lab 5: Vagrant + VirtualBox + QuickNotes; snapshot lifecycle; resource comparison
📖 Read this week:
- 📕 Modern Operating Systems — Andrew Tanenbaum — Chapter 7 (Virtualization)
- 📗 VMware Workstation paper (1999) — how they made x86 virtualization work
- 📘 Vagrant docs — start with "Getting Started"
- 📝 Cloudflare on Heartbleed (2014) — real-time analysis of the attack
🛠️ Tools to install this week: VirtualBox 7.1.x, Vagrant 2.4.x

graph LR
    P["💻 Week 4<br/>OS + Networking"] --> Y["📍 You Are Here<br/>Virtualization"]
    Y --> N["🐳 Week 6<br/>Containers"]
    N --> M["🔧 Week 7<br/>Config Mgmt (Ansible)"]

🎯 Remember: Virtualization was the abstraction that made everything else — clouds, containers, immutable infrastructure — possible. The pattern (run N tenants on shared hardware via a thin abstraction layer) keeps repeating.

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

📌 Lecture 5 — Virtualization: One Box, Many Worlds

📍 Slide 1 – 💥 The Day VMware Saved a Datacenter

📍 Slide 2 – 🎯 Learning Outcomes

📍 Slide 3 – 🗺️ Lecture Overview

📍 Slide 4 – 💾 What Is a Virtual Machine, Really?

📍 Slide 5 – 🏛️ Type 1 vs Type 2 Hypervisors

📍 Slide 6 – 📜 A Short History of x86 Virtualization

📍 Slide 7 – 🛠️ VM Lifecycle: From Image to Running

📍 Slide 8 – 📦 VM Images & Disk Formats

📍 Slide 9 – 📸 Snapshots: The Time Machine

📍 Slide 10 – 🔗 VM Networking: Three Modes

📍 Slide 11 – 📜 Vagrant: Make VMs Reproducible

📍 Slide 12 – 🚀 cloud-init: The Other Half of Reproducibility

📍 Slide 13 – 📈 Resource Sizing: How Much Is Enough?

📍 Slide 14 – 🐳 VM vs Container: The 30-Second Preview

📍 Slide 15 – 🧪 Lab 5 Preview: QuickNotes in a VM

📍 Slide 16 – ❌ Common VM Antipatterns

📍 Slide 17 – 📜 Real Story: Heartbleed and the Patch Storm

📍 Slide 18 – 🧠 Key Takeaways

📍 Slide 19 – 🚀 What's Next + 📚 Resources

Uh oh!

FilesExpand file tree

lec5.md

Latest commit

History

lec5.md

File metadata and controls

📌 Lecture 5 — Virtualization: One Box, Many Worlds

📍 Slide 1 – 💥 The Day VMware Saved a Datacenter

📍 Slide 2 – 🎯 Learning Outcomes

📍 Slide 3 – 🗺️ Lecture Overview

📍 Slide 4 – 💾 What Is a Virtual Machine, Really?

📍 Slide 5 – 🏛️ Type 1 vs Type 2 Hypervisors

📍 Slide 6 – 📜 A Short History of x86 Virtualization

📍 Slide 7 – 🛠️ VM Lifecycle: From Image to Running

📍 Slide 8 – 📦 VM Images & Disk Formats

📍 Slide 9 – 📸 Snapshots: The Time Machine

📍 Slide 10 – 🔗 VM Networking: Three Modes

📍 Slide 11 – 📜 Vagrant: Make VMs Reproducible

📍 Slide 12 – 🚀 cloud-init: The Other Half of Reproducibility

📍 Slide 13 – 📈 Resource Sizing: How Much Is Enough?

📍 Slide 14 – 🐳 VM vs Container: The 30-Second Preview

📍 Slide 15 – 🧪 Lab 5 Preview: QuickNotes in a VM

📍 Slide 16 – ❌ Common VM Antipatterns

📍 Slide 17 – 📜 Real Story: Heartbleed and the Patch Storm

📍 Slide 18 – 🧠 Key Takeaways

📍 Slide 19 – 🚀 What's Next + 📚 Resources