Reliability vs scalability vs performance

These three words get mixed in interviews. Separate them cleanly.

Reliability   -> correct + on time + recoverable when things break
Scalability   -> keeps working as data / users / sources grow
Performance   -> speed / cost efficiency of a given workload

Examples

• Reliability: idempotent loads, tests, alerts, replay path. A slow but correct daily job can still be reliable.
• Scalability: partitioning, horizontal workers, Kafka partitions, warehouse auto-scale. Handles 10× volume without redesign.
• Performance: runtime and resource use for today's 1TB job (prune, avoid shuffle, compact files).

Trade-offs

Chasing micro-latency can hurt reliability (more moving parts). Over-provisioning for performance wastes cost without improving correctness.

         reliability
            / \
           /   \
  scalability---performance

Interview tip: Define each in one line with an example. Then say good pipeline design balances all three against SLA and budget.