Score yourself

Give yourself 0, 1 or 2 points per row.

Area	0 points	1 point	2 points
Requirements	Started drawing	Upload and download	Sync, sharing, history, big files and conflicts agreed
Estimates	None	Numbers, not used later	Exabytes → object storage, rows → sharded metadata, bandwidth → direct uploads
API and data	Skipped	Upload API only	Check-missing, presigned upload, commit with base version, changes cursor
High-level design	Files through app servers	Storage + DB, no sync	Upload, commit, notify and sync walked
Deep dive	Hand-waved	One topic	Chunking with CDC, dedupe, conflicts or metadata sharding done properly
Trade-offs	None named	Named when asked	Raised unprompted (fixed vs CDC, cross-user dedupe privacy, conflict copies)
Communication	Long silences	Mostly clear	Thought out loud, checked in, managed time

12 to 14: strong hire signal. 8 to 11: solid, polish the weak rows. Under 8: redo the round in a week, timer on.

The bar: fresher vs SDE-2

	Fresher	SDE-2
Storage	Files in S3, paths in a DB	Content-addressed blocks, file = list of hashes, versions for free
Transfer	Upload the whole file	Upload only missing blocks, resumable, presigned and multipart
Sync	Poll for changes	Journal + cursor per device, long-poll notifications
Conflicts	Last write wins	Compare-and-set on base version, conflicted copies
Scale	One database	Sharded by namespace, hot folders, GC of unreferenced blocks

Common mistakes

Streaming file bytes through the app servers.
Re-uploading whole files for small edits.
Last write wins, silently losing an offline edit.
Polling every device every few seconds instead of notifications plus a cursor.
Deleting blocks still referenced by older versions or other files.

Quick check

Five questions on this chapter. Pick an answer, then check it.

quiz

1[
2 {
3  "q": "Why split files into hashed blocks?",
4  "options": [
5   "To compress them",
6   "Changed files only upload changed blocks, and identical blocks are stored once",
7   "For encryption",
8   "Storage requires it"
9  ],
10  "answer": 1,
11  "explain": "Blocks by hash, files as lists of hashes, a journal to sync from."
12 },
13 {
14  "q": "Why content-defined chunking instead of fixed-size blocks?",
15  "options": [
16   "Smaller blocks",
17   "Inserting a byte near the start doesn't shift every later block boundary",
18   "It's faster to hash",
19   "No reason"
20  ],
21  "answer": 1,
22  "explain": "Cut by content, not by ruler, and an edit only changes its neighbours."
23 },
24 {
25  "q": "Two devices edit the same file offline. What does a safe sync do?",
26  "options": [
27   "Last write silently wins",
28   "Detect the conflict from the base version and keep both copies",
29   "Delete both",
30   "Lock the file forever"
31  ],
32  "answer": 1,
33  "explain": "Say which version you started from; lose the race, keep both copies."
34 },
35 {
36  "q": "Renaming a 10 GB file should cost...",
37  "options": [
38   "Re-uploading 10 GB",
39   "A small metadata change; no bytes move",
40   "A full copy",
41   "Nothing is possible"
42  ],
43  "answer": 1,
44  "explain": "Shard by folder, transact inside it, never touch bytes for a rename."
45 },
46 {
47  "q": "How do bytes get to storage without passing through your servers?",
48  "options": [
49   "They can't",
50   "Presigned upload URLs straight to object storage",
51   "Through the database",
52   "Email"
53  ],
54  "answer": 1,
55  "explain": "We sign the permission slip; the bytes go straight to storage."
56 }
57]

39. Design File Storage and Sync (Google Drive / Dropbox)

Score yourself

Score yourself

The bar: fresher vs SDE-2

Common mistakes

Quick check

Platform

Company

Legal