Part 23: Case Studies

Linux Overview

Linux is a free, open-source kernel powering everything from phones (Android) to supercomputers.

Linux Timeline:
══════════════════════════════════════════════════════════════

1991  Linus Torvalds releases Linux 0.01
1994  Linux 1.0 (TCP/IP, X Window)
1996  Linux 2.0 (SMP support)
2003  Linux 2.6 (O(1) scheduler)
2011  Linux 3.0 (cgroups mature)
2020  Linux 5.x (io_uring, eBPF)

Design Philosophy:
• "Everything is a file"
• Small tools that do one thing well
• Open development, GPLv2 license
• Performance and flexibility over ease of use

Windows Overview

Windows NT is Microsoft's kernel since Windows NT 3.1 (1993), designed for compatibility and enterprise features.

Windows NT Architecture:
══════════════════════════════════════════════════════════════

 User Mode
 ┌───────────────────────────────────────┐
 │  Win32 Apps | UWP Apps | Subsystems  │
 ├───────────────────────────────────────┤
 │        NTDLL.DLL (NT API)             │
 └───────────────────────────────────────┘
                    │
 Kernel Mode        ▼
 ┌───────────────────────────────────────┐
 │    Executive (NT Kernel Services)     │
 │  • Process Mgr   • Memory Mgr         │
 │  • I/O Mgr       • Object Mgr         │
 ├───────────────────────────────────────┤
 │    HAL (Hardware Abstraction Layer)   │
 └───────────────────────────────────────┘

Key Features:
• Object-based design (everything is an object)
• Strong backward compatibility
• Registry for configuration
• Hybrid kernel (microkernel-like structure, monolithic performance)

macOS Overview

XNU ("X is Not Unix") is Apple's hybrid kernel combining Mach microkernel with BSD Unix.

XNU Kernel Structure:
══════════════════════════════════════════════════════════════

┌───────────────────────────────────────┐
│           BSD Layer (FreeBSD)          │
│   POSIX APIs, VFS, Networking Stack    │
├───────────────────────────────────────┤
│            Mach Microkernel            │
│   IPC, Memory, Thread scheduling       │
├───────────────────────────────────────┤
│              I/O Kit                   │
│   C++ driver framework                 │
└───────────────────────────────────────┘

Unique Features:
• Grand Central Dispatch (GCD) for concurrency
• Mach ports for secure IPC
• Apple Silicon (ARM) unified memory architecture

Kernel Comparison

Kernel Architecture Comparison:
══════════════════════════════════════════════════════════════

Aspect          Linux           Windows NT      macOS (XNU)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Type            Monolithic      Hybrid          Hybrid
Source          Open (GPL)      Closed          Partial open
LOC             ~30M            ~50M (est.)     ~8M

Syscall         syscall inst    sysenter/int2e  mach_msg/BSD
IPC             Pipes, sockets  Named pipes,    Mach ports
                signals, SysV   LPC/ALPC        BSD sockets

Scheduling Comparison

CPU Scheduling Approaches:
══════════════════════════════════════════════════════════════

LINUX (CFS - Completely Fair Scheduler):
• Red-black tree of tasks, sorted by virtual runtime
• Tasks that run less get scheduled more
• Per-CPU runqueues, periodic load balancing
• Priority: nice values (-20 to +19)

WINDOWS (Multilevel Feedback Queue):
• 32 priority levels (0-31)
• Higher priority always preempts lower
• Priority boost for I/O-bound tasks
• Foreground window priority boost

macOS (Mach-based):
• Thread-centric scheduling (not process)
• Quality of Service (QoS) classes
• Grand Central Dispatch integrates with scheduler
• Priority: kernel manages based on QoS

Memory Management

Virtual Memory Implementation:
══════════════════════════════════════════════════════════════

LINUX:
• 4-level page tables (x86-64)
• Demand paging with copy-on-write
• Slab allocator for kernel objects
• OOM killer when memory exhausted

WINDOWS:
• Paged and non-paged pools
• Working Set management (per-process)
• Standby list for quick reclaim
• SuperFetch (predictive loading)

macOS:
• Compressed memory (in-RAM compression)
• Memory pressure notifications to apps
• Unified memory on Apple Silicon (CPU+GPU share)

File System Comparison

Default File Systems:
══════════════════════════════════════════════════════════════

Linux: ext4 (default), btrfs, XFS, ZFS
  • Journaling, extents, delayed allocation
  • Case-sensitive by default
  • POSIX permissions + ACLs

Windows: NTFS (default), ReFS
  • MFT (Master File Table)
  • Case-insensitive (preserving)
  • ACLs with inheritance
  • Alternate data streams

macOS: APFS (Apple File System)
  • Copy-on-write, snapshots
  • Space sharing between volumes
  • Native encryption
  • Case-insensitive by default (macOS)

Max File Size:
  ext4: 16 TiB      NTFS: 16 EiB      APFS: 8 EiB

Conclusion & Next Steps

Each OS represents different trade-offs:

• Linux: Flexibility, transparency, server dominance
• Windows: Compatibility, enterprise features, desktop market
• macOS: Integration, user experience, creative workflows