DEVOURED
The Database at 550 Kilometers: What Orbital Computing Means for Distributed Databases
Running distributed databases in space requires moving from geographic localities to temporal ones as satellites shift their positions relative to ground stations.
What: Cockroach Labs researcher Isaac Wong argues that while space compute is efficient, storing data in orbit remains expensive, suggesting a hybrid model where hot data lives on satellites and cold data on Earth.
Why it matters: This reveals that distributed consensus protocols like Raft are fundamentally ready for dynamic, non-terrestrial topologies, but require smarter, physics-aware leaseholder management to maintain latency bounds.
Deep dive
- Orbits and Data: Low Earth Orbit (LEO) satellites require a 'locality' system that accounts for constant movement (550km altitude at 7.66 km/s).
- Economics: Compute is dense in value, making space-based GPUs viable; storage has high mass and launch costs, making terrestrial persistence more economical.
- Consensus: Consensus (Raft) is viable in space, but latency is variable due to shifting network geometry, not just network congestion.
- Data Sovereignty: Regulations (GDPR) based on physical borders become complex when servers pass over dozens of countries in 95 minutes.
- Physics: Relativistic clock drift is negligible for databases compared to propagation delays; Newtonian physics is the primary constraint.
Decoder
- Leaseholder: In CockroachDB, the node responsible for coordinating reads and writes for a specific range of data.
- Raft: A consensus algorithm used to manage a replicated log, ensuring all nodes in a cluster agree on the order of operations.
- Hybrid Logical Clocks (HLC): A way to order events in a distributed system that combines physical time with logical counters to handle clock skew.
Original article
Full article content is not available for inline reading.
Explore all 68 articles in today's digest
Fresh dev news, AI-digested daily. Never miss a trend at devoured.fyi