RocksDB

2026 RocksDB Meetup at Meta HQ

Mon, Mar 16 · 3:30 PM PDT

Facebook Building 23, 300 Constitution Drive, Menlo Park, CA, US

[Update: You will need to show a visitor pass to security to access our parking garage / campus]

We’re excited to announce our 2026 RocksDB meetup!

This year, our annual RocksDB meetup will take place in the spring, and we look forward to bringing the community together. We’re thrilled to share that both FoundationDB and Cloudflare will present on how they leverage RocksDB in their internal systems. Additionally, RocksDB engineers at Meta will showcase their latest innovations in the storage engine space. Dinner will be provided as well as Swag!

Date: Monday, March 16 from 3:30-8 pm
Where: Menlo Park Building 23, Conference Room: Flubber
For remote attendees, here is the zoom link.

Agenda:
3:30 - 4:10 pm - Check-in/social time
4:10 pm: Kick-off & Welcome
4:15-6:30: RocksDB Talks by Industry, Academia, and internal Meta engineers. See below for talks
6:30 - 8 pm: Social Hour & Dinner

Agenda:

Title: How Cloudflare Built Instant Cache Purge with RocksDB

Abstract: Cloudflare's CDN caches content across 330+ data centers, and when customers need to invalidate that content, it must happen everywhere -- fast. We built CacheDB, a Rust service running a per-machine RocksDB instance on every edge server, to index cached assets by tag, hostname, and URL prefix for instant purge. This architecture delivers sub-150ms global purge latency (P50) -- a 90% improvement over our previous system. This talk covers the RocksDB schema design using five column families with distinct tuning profiles, prefix iteration for flexible purge lookups, a dual lazy+active purge mechanism, managing tombstone churn in a high-delete workload, and production lessons from operating thousands of RocksDB instances at the edge.

RockDB Storage Engine in FoundationDB
Abstract: FoundationDB is a distributed, strictly serializable key–value store designed to provide strong transactional guarantees. It recently adopted RocksDB as an alternative storage engine within its storage server architecture, introducing new interactions between LSM-tree behavior and a latency-sensitive, replicated transaction system. In this talk, we share our experience integrating RocksDB into FoundationDB, highlighting the engineering trade-offs, performance pathologies, and cross-layer design decisions required to achieve predictable performance at scale.

Sally: Meta's Unified Storage Management Plane
Speaker: Dhanabal Ekambaram

Lightning Talks:
RocksDB Compression Revamp - Peter Dillinger
Key-Value Separation in Data Blocks - Josh Kang
Resumable Compaction - Hui Xiao

RocksDB: State of the Storage Engine. Latest features, innovations, and investments in the future.
Speaker: Xingbo Wang

Note: Please check back on this event as we will update with more details as the date approaches.

Hope to see you there!