Why TigerBeetle is the Most Interesting Database in the World TigerBeetle is a cutting-edge database designed specifically for handling financial transactions with unprecedented performance, reliability, and simplicity. It defies conventional database norms by adopting novel engineering principles and state-of-the-art technologies. --- Key Differentiators of TigerBeetle Writes code slowly and carefully, emphasizing quality over speed. Built entirely on Deterministic Simulation Testing (DST) rather than minimal testing. Has zero external dependencies, unlike many software stacks. Employs static memory allocation and keeps assertions enabled even in production. Uses Viewstamped Replication for consensus instead of Raft. Written entirely in Zig, a modern system programming language. --- The Database Designed for Debits and Credits TigerBeetle is called “The Financial Transactions Database,” using debits and credits as its core primitives. Inspired by early database transaction theory by Jim Gray, where business transactions are modelled as debits and credits. Standard SQL databases handle such transactions poorly, requiring many queries and produce issues such as hot rows and locks. TigerBeetle can pack 8,190 debit/credit operations into a single 1MiB query with just one roundtrip. It is ready for the explosive growth in real-time and instant payments worldwide. TigerBeetle has passed Jepsen testing with exceptional robustness, proving its reliability. --- Modern Architecture Distributed by Default Designed to run on cloud infrastructure, with strict serializable replication across multiple nodes. Offers simple deployment with no complex external systems required. Implements Viewstamped Replication consensus from MIT, designed for fault tolerance. Clock Fault Tolerance Uses a fault-tolerant "cluster time" by combining multiple system clocks and applying sophisticated algorithms (e.g., Marzullo’s algorithm) to detect and handle clock skew and faults. Alerts operators if clock synchronization protocols like NTP fail. Storage Fault Tolerance Unlike traditional databases that assume reliable disk behavior, TigerBeetle: Uses Protocol Aware Recovery to maintain availability unless all replicas of data are corrupted. Stores data as immutable, checksummed, and hash-chained to prevent corruption or tampering. Writes directly to disks with minimal software layers, including support for raw block devices. Implements its own custom multi-tree LSM Forest storage engine. --- Why Zig? Combines the power of C’s ecosystem with modern tooling and readability. Supports static memory allocation, critical for TigerBeetle's predictability. Easier developer experience leading to faster onboarding. Some TigerBeetle team members were involved in Rust development but chose Zig for project-specific reasons. --- Deterministic Simulation Testing (DST) and VOPR DST runs a simulator that exhaustively explores all concurrency scenarios in a distributed system on a single thread, testing various faults and delays. TigerBeetle operates the largest known DST cluster called VOPR (Viewstamped Operation Replicator): Uses 1,000 CPU cores running 24/7 to simulate nearly 2 millennia of cluster runtime every day. There is even an in-browser game simulating TigerBeetle’s behavior built on WebAssembly, making DST accessible and fun. --- Engineering Ethos: TigerStyle and The Power of Ten TigerStyle is TigerBeetle’s public coding and engineering methodology blending art and science. Inspired by NASA’s Power of Ten, it emphasizes: Extensive use of assertions for all inputs, outputs, conditions, and invariants to catch bugs early. Prioritizing code reasoning and design over just writing code, especially for performance. Early performance considerations using