May 25, 2026
Gorilla Warfare in the Comments
Gorilla: A fast, scalable, in-memory time series database (2016)
Facebook’s data beast saved the day — but the comments instantly started a nerd fight
TLDR: Facebook built Gorilla to catch site problems almost instantly while handling a mind-bending flood of system data. Commenters were impressed, but the thread quickly turned into a familiar tech showdown: one person pushed a rival tool, while another asked the internet’s favorite question — why not compare it to DuckDB?
Facebook’s old-school monitoring hero Gorilla is being remembered as the beast that caught trouble fast: when errors started exploding on one of the company’s sites, this system spotted the problem within minutes and helped engineers trace it back to something as boring-looking as copying a software file. That’s the kind of story tech people love — a giant invisible safety net quietly stopping chaos before users notice. And the scale is the real jaw-dropper: billions of tracking streams, trillions of measurements a day, and answers coming back in under a blink.
But of course, the community didn’t just clap politely. The real action in the comments was the classic internet sport of “cool, but have you tried my favorite tool?” One commenter basically gave Gorilla a compliment sandwich before sliding in a rival suggestion, pointing to Sprintz as the smarter way to squeeze the data down even more. Translation for normal people: “Nice giant machine you built, but I think I know a cheaper trick.” Then another commenter dropped the wonderfully chaotic, ultra-2020s question: “How does it compare to DuckDB?” And just like that, the thread swerved from admiration into brand-comparison cage match territory.
The vibe? Half awe, half backseat driving. People seem impressed that Facebook built a monster system that lives mostly in memory for speed, but commenters also can’t resist turning every engineering win into a debate, a recommendation thread, and a low-key meme about comparing every database on Earth to DuckDB.
Key Points
- •Facebook used Gorilla, an in-memory time series database, to detect and help diagnose production issues, including incidents identified through automated alerts and metric correlation.
- •As of spring 2015, Facebook’s monitoring systems generated more than 2 billion unique time series and about 12 million data points per second, exceeding 1 trillion data points per day.
- •Gorilla’s design goals included retaining 26 hours of data in memory, supporting up to 40,000 queries per second, delivering reads in under 1 millisecond, and surviving crashes with replicated in-memory data.
- •Gorilla functioned as an in-memory write-through cache backed by HBase and used a compression algorithm that achieved an average 12x reduction in size.
- •Its horizontally scalable, share-nothing architecture grew from 20 machines holding 1.3 TB of RAM at launch to 80 machines per cluster as Facebook’s monitoring data expanded.