Database Comparison: SQLite vs. PostgreSQL vs. MongoDB

Alright, picking a database for QuickLinker-ZeroRPM was a big damn deal—I needed something that could handle 100,000 requests a minute, cost me nothing, and not fall apart under pressure. I messed around with SQLite, PostgreSQL, and MongoDB, and here’s the straight dope on why SQLite’s my champ, PostgreSQL’s a solid starter, and MongoDB’s out of the running. Let’s break it down.

SQLite: The Scrappy Winner

This little beast is why I’m hitting my goals—cheap, fast, and tough when paired with Redis.

• Cost: $0. It’s just a file—no server, no fees. Lives on Oracle’s free tier (4 cores, 24 GB RAM) without blinking.
• Speed: Tuned it up—WAL mode, sync tweaks—and it’s cranking 80,000 inserts a second (SQLite tuning). At 100,000 RPM (1,666 req/s), with 1% writes (16.7/s) and 99% reads (1,650/s), Redis grabs 99% of reads. SQLite’s left with ~33 ops/s—pocket change.
• Fit: Low writes, high reads? That’s URL shortening in a nutshell—bit.ly’s 230 writes/min at 23,000 RPM proves it (Bitly stats). SQLite eats that up.
• Upside: LiteFS in Phase 3 adds HA for free—keeps it bulletproof (LiteFS GitHub).

I love it ‘cause it’s lean and mean—keeps my $0 dream alive and scales with some grit.

PostgreSQL: The Reliable Rookie

I kicked off Phase 1 with PostgreSQL—it’s got chops, but it’s not the endgame.

• Cost: $0 on the free VM for 100 req/s, but scaling to 100,000 RPM? Read replicas kick in at $10-20 a month—like AWS RDS territory.
• Speed: Handles 100 req/s no sweat—ACID-compliant, solid for my short_code → original_url setup. But 1,666 req/s needs more juice—connection pooling and replicas eat CPU and RAM.
• Fit: Great for Phase 1’s small scale—consistent, simple, no duplicates. Switching to SQLite later was easy since they speak the same SQL lingo.
• Downside: It’s a server hog. More RAM, more overhead—fine ‘til I hit big traffic, then it’s cash or bust.

It’s my trusty starter, but it’s too thirsty for the long haul.

MongoDB: The Fancy Flop

MongoDB sounded cool—big name, big promises—but it’s a mismatch for my gig.

• Cost: $0 on one VM, but that’s sketchy for production. Real HA needs a 3-node cluster—$15-30 a month, like MongoDB Atlas at $5 a node. No way I’m paying that.
• Speed: Loves reads, sure—scales like crazy there. But Redis already owns 99% of my reads (1,633/s). Mongo’s stuck with writes, and its eventual consistency could screw me—two threads grabbing the same short_code? Disaster.
• Fit: Overkill for short_code → original_url. Its document flex is wasted here, and moving to SQLite later would’ve been a nightmare—no SQL overlap.
• Downside: Too fat—3 nodes need 3-6 GB RAM and heavy I/O. My free VM chokes on that.

Mongo’s a beast, but not my beast—Redis steals its thunder, and the price tag’s a dealbreaker.

Head-to-Head: Who’s Got What

Database	Cost (100,000 RPM)	Speed Fit	Resource Load	HA at $0?
SQLite	$0 (LiteFS)	80,000 ops/s	Light as hell	Yup
PostgreSQL	$10-20/month	Good ‘til 100 req/s	Chunky server	Nope
MongoDB	$15-30/month	Read-heavy king	Fat and greedy	Nope

Takeaway: SQLite’s the only one that hits 100,000 RPM at $0—fast, light, and future-proof with LiteFS.

Why SQLite Wins

• Cash Rules: $0 from day one to 100,000 RPM. PostgreSQL and MongoDB tap out when I scale.
• Workload Match: Low writes (16.7/s), high reads (1,650/s)—Redis and SQLite split it perfect. Mongo’s read edge is moot, and PostgreSQL’s too bulky.
• Flex: LiteFS makes SQLite HA-ready; the others need paid crutches.

I started with PostgreSQL ‘cause it’s safe and simple—got me to 100 req/s no fuss. But SQLite’s where I landed—same SQL vibe, way cheaper, and with Redis, it’s untouchable. MongoDB? Cool for someone else’s wallet.

What I Learned

• SQLite’s Sneaky Good: Thought it was small-time, but it’s a tank with the right setup.
• Cost Kills: PostgreSQL and MongoDB are dope ‘til the bill hits—$0’s my line in the sand.
• Teamwork Wins: SQLite + Redis + LiteFS = my dream squad.

What’s Next?

• Phase 1: Where PostgreSQL hands off to SQLite.
• SQLite + LiteFS: Deep dive on the duo.
• References: The dirt I dug up.

---

Last Updated: March 11, 2025