Insights:How to Choose Monitoring & Alerting by Company Scale
Scale One: Single-Site Small Project (1–3 sites)
Requirements: Know whether the site is down, certificates are expiring, and forms can be submitted.
| Item | Recommendation |
|---|---|
| Detection Method | Free or low-cost external probing, 1–5 minutes |
| Probe Locations | 2 regions is enough |
| Alert Channel | IM group + email daily report |
| On-Call | 1 person + 1 backup |
| Cost | Under a few hundred yuan per year, or free tier allowance |
Key point: Don't build your own monitoring platform. Use small tools to cover "site unreachable" and "certificate nearing expiry" and that's sufficient.
Scale Two: Multi-Site Medium (4–20 sites, with dedicated ops)
Requirements: Unified dashboard, graded alerts, avoid duplicate configuration.
| Item | Recommendation |
|---|---|
| Detection Method | Unified platform, bulk import sites |
| Probe Locations | 3–5 regions |
| Alert Grading | P1/P2/P3 three-tier |
| On-Call | Schedule + escalation rules |
| Cost | Thousand-yuan-level tool fees per year + labor |
Key point: Prioritize resolving "configuration consistency" and "alert deduplication." Multi-site setups most easily trigger dozens of alerts for the same incident.
Scale III: Multi-business Enterprise (20+ Sites or Critical Transaction Systems)
Requirements: Business-level availability, dependency chain tracing, SLA analytics.
| Item | Recommendation |
|---|---|
| Detection Method | External probing + Internal metrics + Synthetic transactions |
| Detection Points | Multiple regions and carriers |
| Alert Tiers | Linked to SLA with escalation matrix |
| On-Call | 7×24 rotation or outsourced |
| Cost | Yearly cost in ten-thousands range; requires dedicated staff |
Key Point: At this scale, monitoring itself becomes a system. Someone must own rule governance, or within six months nobody will be checking alerts anymore.
Comparison of Three Tiers
| Dimension | Small Project | Medium | Large |
|---|---|---|---|
| Core Objective | Detect outages | Unify and deduplicate | Locate issues and track SLA |
| Tool Form | SaaS probing | Monitoring platform | Multi-system integration |
| Human Effort | Part-time | Semi-dedicated | Dedicated |
| Over-engineering Risk | Low | Medium | High |
Questions to Ask During Selection
- When a site goes down, how quickly will someone know?
If you can’t answer these four questions, your approach isn’t ready.
Boundaries
A monitoring plan improves “detection and response speed” but doesn’t change the underlying stability of the system itself. True availability also depends on architecture redundancy, release processes, and capacity planning.
Related Reading: