Midterm 3 Cheatsheet (SP26)

Content locked

The catalog is public. Reading this content requires approved access.

1. I/O Devices (lec 16)

Three protocols

Modern queueing (NVMe-style)

Common traps

# polls in basic PIO (no interrupts, no DMA)

2. HDDs (lec 16)

Anatomy (terms used in lecture)

Timing formula

Disk performance implications

Common traps

3. SSDs (lec 16/17)

Architecture

FTL (Flash Translation Layer)

Garbage collection & overprovisioning

Common traps

4. RAID (lec 23)

General framework: mapping + redundancy

Levels (N disks total)

Parity math

Mapping (block size = chunk size; lecture assumes chunk size = 1)

Write costs (RAID 4/5)

Consistent-update problem (RAID 1)

Common traps

5. File System API (lec 17)

File descriptor (fd)

Three layers of state

Key calls

Atomic file update pattern (lec 17 — important)

Common traps

6. VSFS (lec 19)

On-disk layout

Inode contents

Pointer reach (4 KB blocks, 4 B pointers → 1024 ptrs/block)

Directories

Refcount math

Hard link vs symlink

Path resolution cost

7. FFS (lec 20)

Idea

Allocation policies

Other FFS features (from lec 20)

Common traps

8. LFS (lec 21)

Idea

Write path

Lookup chain

Imap

Garbage collection

Segment Summary (SS)

Crash recovery

Write amplification

Common traps

8.5. File System Family Comparison

Allocation strategies (lec 19)

File system designs (lec 19–22)

Quick decision rules (often tested)

9. Crash Consistency / Journaling (lec 22)

The problem

Inconsistencies (per pair written before crash)

Consistency vs Atomicity (lec 22 distinction)

fsck (Solution #1)

Journaling lifecycle

Recovery rule

Modes (lecture terminology)

Barriers

Common traps

10. Containers (lec 24)

What it is

Three primitives

Container vs VM

Container images

Common traps

11. Distributed Systems / NFS (lec 25)

Distributed challenges

Reliable messaging building blocks (lec 25)

NFS architecture

Protocol design evolution (lecture: strategies 1–5)

Idempotency (the crux)

Server crash

Caching & consistency

Write buffering (server side)

Common traps

12. Worked Examples & Past-Exam Traps

A. Disk timing arithmetic

B. Tradeoff intuition (often T/F)

C. RAID failure semantics

D. Even parity calculation

E. RAID-6 (mentioned in sample-1)

F. RAID write/read tradeoffs

G. Path resolution inode count

H. Disk writes for a basic create+write

I. Multi-level indexed allocation max file size

J. Inode does NOT contain

K. Hard link vs symlink + rename (sample-3 Q18-21)

L. Crash without journaling (sample-3 Q22-27 master pattern)

M. Journaling write count → recovery state (sample-3 Q28-34)

N. Journal bitmap (sample-1 Q55-56)

O. fsync / atomic file update

P. fork / dup / open offset semantics

Q. FFS placement edge cases

R. LFS subtleties

S. Direct-mapped SSD random writes (sample-1 Q44, sample-2 Q49)

T. NFS scenarios (sample-2 Q53-56 archetypes)

U. NFS idempotency catalog

V. TCP vs UDP

13. Past-Exam Question Bank (in-scope only)

I/O Devices

HDDs (timing only — scheduling out of scope)

SSDs

RAID

File API (fd / fork / dup)

VSFS / Inodes

FFS

LFS

Crash Consistency / Journaling

NFS / Distributed

Document