DEV Community

Cover image for Self-Hosted Observability: $20/month vs $800/month
Polliog
Polliog

Posted on

Self-Hosted Observability: $20/month vs $800/month

I run Logtide's production environment, handling 500k logs/day, on a €5.49/month Hetzner server.

Datadog quoted me $800/month for the same workload.

Here's the honest cost breakdown.

The SaaS Quote (Real Numbers)

I plugged our actual usage into Datadog's pricing calculator:

What we send:

  • 500k logs/day = ~15GB/day
  • 10 services with APM traces
  • 30-day log retention
  • 1 server (we run everything on one box)

Datadog's quote:

  • Log ingestion: 15GB/day × $0.10 = $45/month (just to collect)
  • Log indexing: ~250M events × $1.70/million = $425/month (to make them searchable)
  • APM: 1 host × $31/month = $31/month
  • Infrastructure monitoring: 1 host × $15/month = $15/month

Total: $516/month just for basics.

Add retention beyond 15 days? Another $100-200/month.
Want custom metrics? More money.
Need security monitoring? Keep adding.

Real-world Datadog bills easily hit $800-1,200/month for this workload.

The Self-Hosted Stack

What you actually need:

  • Logtide (logs + traces + SIEM)
  • Prometheus (metrics)
  • Grafana (dashboards)

Infrastructure (Hetzner CX33):

  • 4 vCPUs, 8GB RAM, 80GB SSD
  • €5.49/month ($6/month)
  • 20TB bandwidth included
  • EU datacenter (GDPR-compliant)

Backups:

  • Hetzner snapshots: €1/month
  • Off-site to Backblaze B2: ~$2/month

Total infrastructure: $9/month

Add a domain ($12/year = $1/month) and SSL (Let's Encrypt = free).

All-in cost: ~$10-20/month depending on backup strategy.

What We Actually Run

Our production setup on that €5.49 server:

  • Logtide (PostgreSQL + TimescaleDB)
  • Prometheus
  • Grafana
  • The actual application (5 microservices)
  • Redis (for now, making it optional soon)

Resource usage:

  • CPU: 30-40% average
  • RAM: 6GB / 8GB used
  • Disk: 25GB / 80GB used (with TimescaleDB compression)

Performance:

  • Log query latency: 50-150ms
  • Handles 10k logs/sec bursts
  • 30-day retention = 15GB total (compressed from ~150GB raw)

It just works.

The Hidden Cost: Time

Be honest about this part.

Initial setup:

  • Learning curve: 4-8 hours (if new to self-hosting)
  • Initial deployment: 2-4 hours
  • Dashboard setup: 2-3 hours

Total setup: ~10 hours one-time

Monthly maintenance:

  • Updates/patches: 30 minutes
  • Monitoring the monitoring: 15 minutes
  • Backup verification: 15 minutes

Total ongoing: ~1 hour/month

If your time is worth $100/hour:

  • Setup cost: $1,000 (one-time)
  • Monthly cost: $100

First year total:

  • Infrastructure: $120
  • Time: $1,000 (setup) + $1,200 (ongoing)
  • Total: $2,320

Still way cheaper than Datadog ($9,600/year minimum).

Year two and beyond:

  • Infrastructure: $120
  • Time: $1,200
  • Total: $1,320/year

The time cost drops significantly after setup.

What You Lose

Let's be honest about the downsides.

No fancy UI out-of-the-box:

  • Grafana is powerful but not as polished as Datadog
  • You build your own dashboards
  • Learning curve exists

No phone support:

  • When it breaks at 2am, you fix it
  • Or you wait until morning (if non-critical)
  • Stack Overflow and GitHub are your friends

You manage uptime:

  • Datadog has 99.9% SLA
  • Self-hosted: you're responsible
  • (But honestly, Hetzner's uptime is solid)

Scaling is manual:

  • Datadog scales automatically
  • You add more servers yourself
  • Not an issue until you hit millions of logs/day

What You Gain

Data ownership:

  • Logs never leave your server
  • GDPR-compliant by default
  • No third-party data processing

Predictable costs:

  • $20/month whether you send 100k or 500k logs/day
  • No surprise bills
  • Sleep better

No vendor lock-in:

  • It's just PostgreSQL
  • Export anytime
  • Switch tools freely

Learning:

  • You actually understand your stack
  • Better debugging skills
  • Your team learns infrastructure

Real Production Numbers

From our deployment:

Storage with TimescaleDB compression:

  • Raw logs: ~150GB/month
  • Compressed: ~15GB/month
  • 90% reduction

Query performance:

  • 50-150ms for most queries (10M+ logs)
  • Full-text search with pg_trgm
  • Fast enough for incident response

Costs over 6 months:

  • Hetzner: €32.94 ($36)
  • Backups: ~$12
  • Domain: $6
  • Total: $54

Datadog for same period: ~$4,800

Savings: $4,746

When SaaS Makes Sense

Self-hosting isn't always the answer.

Use Datadog if:

  • Team < 5 engineers (time is precious)
  • Raised funding (cost isn't the bottleneck)
  • Zero DevOps skills in-house
  • Need vendor SLA for compliance
  • Rapid unpredictable scaling

Self-host if:

  • Bootstrapped/pre-revenue
  • Team has basic DevOps skills
  • GDPR/privacy requirements
  • Predictable workload
  • Want full control

Migration Path

If you're on Datadog and curious:

Week 1: Run both in parallel

  • Send logs to Logtide + Datadog
  • Compare dashboards
  • Verify all alerts work

Week 2: Validation

  • Test incident response workflow
  • Check query performance
  • Train team on new UI

Week 3: Cut over

  • Stop sending to Datadog
  • Cancel subscription
  • Monitor for issues

Week 4: Optimize

  • Fine-tune retention
  • Build custom dashboards
  • Document runbooks

Total migration time: ~16 hours spread over 4 weeks.

The Realistic Take

For most small-to-medium deployments, self-hosted wins on cost.

The math:

  • Datadog: $9,600/year minimum
  • Self-hosted: $240/year + setup time
  • Savings: $9,360/year

Even if setup takes you 40 hours at $100/hour ($4,000), you break even in 6 months.

Every year after that is pure savings.

But: If you're a 50-person team with unlimited budget, Datadog's polish might be worth it. The UI is genuinely great. The integrations are extensive. The support is responsive.

For everyone else? A $20/month VPS does the job.


Running Logtide in production? Self-hosting your observability? Drop your setup and costs below.

Curious about the setup?


Disclaimer: I build Logtide. These are real numbers from our production environment. Your mileage may vary depending on log volume and retention needs.

Disclaimer n2: Some prices/statistics can change by your setup or just bt changes of the provider or currency

Top comments (0)