SpacetimeDB team explores ACID databases for real-time application state management

The interview revisits a long-standing architectural assumption: that ACID-compliant databases are unsuitable for real-time application state management. This separation has led to dual-system designs where databases handle persistent or financial data while custom in-memory structures manage runtime state. The discussion suggests this division may not be necessary, though the material does not provide performance benchmarks or concrete implementation details to validate the claim

If ACID databases can meet the latency and throughput demands of real-time applications, it could eliminate the need for separate state management layers. This would reduce code complexity and operational overhead, particularly in applications like multiplayer games where world state must be synchronized across clients. However, the material does not specify the trade-offs involved, such as increased memory usage or the impact on transaction isolation levels

The framing of this as an open question rather than a solved problem is notable. It implies that the team is either in early research or deliberately avoiding overstatement. Without evidence of real-world deployment or comparative testing, the discussion remains theoretical. Engineers evaluating this approach would need to assess whether the benefits outweigh the risks of deviating from established patterns

The lack of corroborating sources or independent testing in the material limits its immediate applicability. For production systems, the burden of proof would fall on demonstrating that ACID databases can match or exceed the performance of specialized in-memory solutions. The interview does not address failure modes, such as how the system behaves under network partitions or high contention scenarios