Operate
- Start from access patterns, shard-aware performance, and failure semantics.
- Treat ScyllaDB as a distributed system with explicit operational tradeoffs, not just "faster Cassandra".
- Prefer simple, workload-driven models and predictable maintenance behavior.
- Optimize for low-latency consistency with operational realism.
Default Standards
- Partition design must fit query and hotspot reality.
- Shard-aware performance should be understood, not assumed.
- Consistency and topology choices must match business semantics.
- Compaction, repair, and maintenance require proactive ownership.
- Capacity planning must include node loss and maintenance tax.
References
- Data modeling and shard-aware partitioning: references/data-modeling-and-shard-aware-partitioning.md
- Consistency, topology, and multi-DC design: references/consistency-topology-and-multi-dc-design.md
- Low-latency performance and workload tuning: references/low-latency-performance-and-workload-tuning.md
- Compaction, tombstones, and storage management: references/compaction-tombstones-and-storage-management.md
- Repair, maintenance, and operational discipline: references/repair-maintenance-and-operational-discipline.md
- Capacity, tenancy, and cluster governance: references/capacity-tenancy-and-cluster-governance.md
- Reliability and operations: references/reliability-and-operations.md
- Incident runbooks: references/incident-runbooks.md