Observability is the practice of understanding a system’s internal state through its external outputs — metrics, logs, and traces. Grafana (open-source, self-hosted or Grafana Cloud) and Datadog (fully managed SaaS) are the two most popular observability platforms in 2026. For Indian development teams, the choice depends on budget, team size, and whether you prefer managing infrastructure or paying for convenience, with costs ranging from ₹0 (Grafana OSS) to ₹2–10 lakh/month (Datadog enterprise).
Last updated October 2026
What Is Observability and Why Is It Different from Monitoring?
Traditional monitoring tells you when something breaks — CPU hit 95%, disk is full, the site returned a 500 error. Observability goes deeper: it helps you understand why something broke, even for failure modes you didn’t predict. The three pillars of observability are:
- Metrics: Numerical time-series data — request rate, error rate, latency percentiles (p50, p95, p99), CPU/memory usage
- Logs: Timestamped text records of events — application errors, user actions, system events
- Traces: End-to-end request paths across services — showing exactly which microservice or database query caused a slowdown
For Indian startups running microservices on AWS Mumbai or GCP Mumbai regions, observability is critical because distributed systems fail in unpredictable ways. A payment failure on Razorpay integration might involve your API gateway, authentication service, order service, payment service, and webhook handler — traces connect these dots.
Grafana Stack: The Open-Source Approach
The Grafana observability stack (often called LGTM) consists of four components:
- Loki: Log aggregation (like a lightweight ELK alternative)
- Grafana: Visualisation and dashboarding
- Tempo: Distributed tracing
- Mimir: Long-term metrics storage (Prometheus-compatible)
Setting Up Grafana for a Node.js Application
For a typical Indian SaaS application built with Node.js and Express, here’s the instrumentation approach:
- Use
prom-clientto expose Prometheus metrics (request duration, active connections, business metrics like orders/minute) - Use
winstonorpinofor structured JSON logging, shipped to Loki via Promtail or the Grafana Agent - Use OpenTelemetry SDK for distributed tracing, exported to Tempo
- Grafana dashboards pull from all three sources for unified views
Grafana Cloud Pricing for Indian Teams
Grafana Cloud offers a generous free tier that suits small Indian startups:
- Free: 10,000 metrics series, 50GB logs, 50GB traces per month
- Pro: starts at ~$29/month (~₹2,400/month) with higher limits
- Self-hosted: ₹0 for software, but you pay for the EC2/VM instances (typically ₹5,000–₹20,000/month for a small-medium setup on AWS Mumbai)
Datadog: The Managed SaaS Approach
Datadog provides a unified platform where metrics, logs, traces, and more (RUM, security, CI visibility) are all integrated out of the box. There’s no infrastructure to manage — install the Datadog Agent and data flows automatically.
Key Datadog Features for Web Applications
- APM (Application Performance Monitoring): Auto-instrumented traces for Node.js, Python, Java, Go, Ruby, .NET
- Real User Monitoring (RUM): Frontend performance from actual users’ browsers — critical for understanding Indian user experience across varied devices and networks
- Synthetic Monitoring: Scheduled checks from global locations (though Indian-region synthetics have limited PoPs)
- Log Management: Ingest, index, and query logs with powerful filtering
- Dashboards: Drag-and-drop creation with 700+ integrations
Datadog Pricing Considerations
Datadog’s pricing is per-host and per-feature, which can escalate quickly:
| Feature | Price (approx. INR/month) |
|---|---|
| Infrastructure (per host) | ₹1,500–₹1,900 |
| APM (per host) | ₹2,600–₹3,300 |
| Log Management (per GB ingested) | ₹8–₹15 |
| RUM (per 10K sessions) | ₹150–₹200 |
| Synthetic Monitoring (per 10K tests) | ₹400–₹600 |
A mid-sized Indian SaaS company with 10 hosts, APM, logs (100GB/month), and RUM can easily hit ₹1.5–3 lakh/month on Datadog. This is why many Indian teams start with Grafana and consider Datadog only when the team scaling justifies the cost.
Grafana vs Datadog: Head-to-Head Comparison
| Aspect | Grafana Stack | Datadog |
|---|---|---|
| Cost for 5-person startup | ₹0–₹5,000/month | ₹40,000–₹80,000/month |
| Setup complexity | Medium-High (self-hosted) / Low (Cloud) | Low |
| Vendor lock-in | None (open-source, Prometheus-compatible) | High (proprietary query language, integrations) |
| Dashboard quality | Excellent (highly customisable) | Excellent (easier out-of-box) |
| Alerting | Good (Grafana Alerting, OnCall) | Excellent (anomaly detection, forecasting) |
| Indian support | Community + Grafana Labs (limited India presence) | Sales team in India, enterprise support |
| Compliance (DPDP) | Full control (self-hosted) | Data leaves India (US/EU datacentres) |
Building Effective Alerts: Avoiding Alert Fatigue
One of the biggest observability mistakes Indian teams make is creating too many alerts. Follow these principles:
- Alert on symptoms, not causes: Alert on “5xx error rate > 1% for 5 minutes” rather than “CPU > 80%”
- Use severity levels: P1 (pages the on-call engineer at 3 AM), P2 (Slack notification during business hours), P3 (weekly review)
- Set meaningful thresholds: Base them on SLOs (Service Level Objectives), not arbitrary numbers
- Route intelligently: Payment failures page the payments team, not the frontend team. Use PagerDuty or Grafana OnCall for routing
Observability for WordPress and PHP Applications
If you’re running a high-traffic WordPress site (common for Indian media and e-commerce), observability looks different:
- New Relic or Datadog APM: PHP agent for tracking slow queries, plugin performance, and WooCommerce checkout bottlenecks
- Query Monitor plugin: Development-time observability for WordPress (free, but not production monitoring)
- Server-level monitoring: Grafana + Prometheus node_exporter for LAMP/LEMP stack metrics
- Uptime monitoring: UptimeRobot (free tier) or Better Uptime for Indian-region checks
Implementing OpenTelemetry: The Vendor-Neutral Standard
OpenTelemetry (OTel) is the CNCF standard for instrumentation. It lets you collect metrics, logs, and traces and send them to any backend — Grafana, Datadog, New Relic, or Jaeger. This prevents vendor lock-in and is the recommended approach for new Indian projects in 2026.
The OpenTelemetry Collector acts as a pipeline: receive data from your application, process it (filter, batch, enrich), and export to one or more backends. You can even send traces to Grafana Tempo and metrics to Datadog simultaneously.
Frequently Asked Questions
Which is better for an Indian startup — Grafana or Datadog?
For early-stage Indian startups (seed to Series A), Grafana Cloud’s free tier or self-hosted Grafana provides excellent observability at minimal cost. Switch to Datadog when your team exceeds 15–20 engineers and the operational overhead of managing Grafana infrastructure becomes more expensive than Datadog’s SaaS pricing.
Does Datadog have servers in India?
As of October 2026, Datadog’s primary data centres are in the US and EU. Data from Indian applications is sent to these regions, which raises DPDP Act compliance considerations for companies handling personal data. Grafana self-hosted on AWS Mumbai or Azure Central India keeps data within the country.
What is the minimum observability setup for a production web application?
At minimum, implement: application logs shipped to a searchable store (Loki, CloudWatch, or Datadog), uptime monitoring with alerting (UptimeRobot free tier works), and error tracking (Sentry free tier). Add APM traces when you have microservices. Total cost can be ₹0 using free tiers.
How much log storage should an Indian SaaS company budget for?
A typical Indian SaaS with 10,000 daily active users generates 5–20GB of logs per day. At Datadog’s pricing, that’s ₹1,200–₹9,000/day just for log ingestion. With Grafana Loki on S3 (Mumbai region), storage costs drop to ₹200–₹800/day. Implement log sampling and retention policies to control costs.
Can I use both Grafana and Datadog together?
Yes. A common pattern is using Grafana as the dashboard layer (it has a native Datadog data source plugin) while running Datadog APM for traces. This gives you Grafana’s superior customisation with Datadog’s auto-instrumentation. However, this adds complexity and cost — recommended only for teams with dedicated platform/SRE engineers.
