Calculator Inputs
Large screens show three columns, smaller screens collapse automatically.Example Data Table
These sample workloads show how queue size, throughput balance, and latency can change projected lag.
| Scenario | Queue MB | Write MB/s | Apply MB/s | Estimated Lag | Status |
|---|---|---|---|---|---|
| 1 | 120 | 14.00 | 22.00 | 11.76 sec | Healthy |
| 2 | 380 | 26.00 | 21.00 | 22.20 sec | At Risk |
| 3 | 90 | 9.00 | 15.00 | 22.59 sec | Healthy |
Formula Used
The estimator combines queue backlog, live write pressure, replica throughput, transport delay, and batching overhead.
Worker Efficiency = min(1.35, 0.80 + (Parallel Workers - 1) × 0.06)
Effective Write Rate = Primary Write Rate × Burst Multiplier
Effective Apply Rate = Replica Apply Rate × Worker Efficiency
Catch-up Time = Pending Queue ÷ (Effective Apply Rate - Effective Write Rate), when positive
Transport Delay = (Network RTT + Serialization Delay) ÷ 1000
Batch Delay = ((Commit Batch Size × Average Transaction Size) ÷ 1024) ÷ Effective Apply Rate
Estimated Lag = (Catch-up Time or Queue-only Lag + Transport Delay + Batch Delay) × (1 + Safety Buffer%)
When apply capacity is lower than effective write pressure, the backlog is treated as growing and the page flags the estimate as higher risk.
How to Use This Calculator
- Enter the current replication queue waiting to be applied.
- Add primary write throughput and the replica's real apply capacity.
- Include network and serialization delays for transport overhead.
- Set workers, batch size, average transaction size, and burst pressure.
- Choose a safety buffer and target SLA in seconds.
- Press Estimate Lag to show the result above the form, then export the result as CSV or PDF.
Frequently Asked Questions
1. What does replication lag mean here?
It is the estimated delay between committed source changes and when the replica becomes consistent, after queue, transport, batching, and safety adjustments.
2. Why use a burst multiplier?
Write traffic often spikes above averages. The burst multiplier stress-tests the estimate so you can model peak pressure instead of calm periods only.
3. What happens when apply rate is lower than write rate?
The queue keeps growing. The calculator marks that case as risky or critical and does not claim a true catch-up time under current conditions.
4. Why are parallel workers capped in effect?
Extra workers help, but gains are rarely perfectly linear. The efficiency cap avoids unrealistic projections when contention, locks, or coordination overhead appear.
5. Is the result suitable for every database engine?
No. It is a planning estimator for common replication behavior. Engine-specific internals, lock waits, disk stalls, and failover rules can change actual lag.
6. How should I choose the safety buffer?
Use a higher buffer when workloads are bursty, networks fluctuate, or measurements are noisy. Stable systems can usually work with smaller margins.
7. What does the risk score summarize?
It blends queue-only lag, throughput imbalance, transport delay, and safety margin into one simple severity signal for quick prioritization.
8. Can I export and share the estimate?
Yes. After calculating, use the CSV button for structured data or the PDF button for a shareable report snapshot.