1 min lesson
Backpressure and partitioning
Use two rows in "Backpressure and partitioning" to state the practical decision rules.
Step 1 of 2
Backpressure and partitioningshed load deliberately; isolate noisy neighbors
When consumers can't keep up, the system must push back rather than melt. Backpressure means the queue depth and rate limits signal producers to slow down or shed low-priority work, so a flood of events doesn't cascade into auth, agent backend and the rest of the shared layer. Partitioning by tenant or key gives you horizontal scale and keeps one heavy customer from drowning everyone else.
- Lever
- Per-consumer concurrency cap
- What it protects against
- A slow endpoint monopolizing workers
- Trade-off
- Caps max throughput to that endpoint
- Lever
- Backpressure / load shedding
- What it protects against
- Cascading overload into the rest of Core Services
- Trade-off
- Some low-priority events delayed or dropped on purpose
- Lever
- Partition by tenant/key
- What it protects against
- Noisy-neighbor starvation; uneven load
- Trade-off
- Hot partitions still need rebalancing
- Lever
- Rate limit per source
- What it protects against
- A runaway producer flooding the pipeline
- Trade-off
- A legitimate burst may get throttled
| Lever | What it protects against | Trade-off |
|---|---|---|
| Per-consumer concurrency cap | A slow endpoint monopolizing workers | Caps max throughput to that endpoint |
| Backpressure / load shedding | Cascading overload into the rest of Core Services | Some low-priority events delayed or dropped on purpose |
| Partition by tenant/key | Noisy-neighbor starvation; uneven load | Hot partitions still need rebalancing |
| Rate limit per source | A runaway producer flooding the pipeline | A legitimate burst may get throttled |
Each lever trades raw throughput for isolation. At 1M+ DAU, isolation is usually the better buy.