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?
- Logtide repo: https://github.com/logtide-dev/logtide
- Docs: https://logtide.dev/docs
- Try cloud version: https://logtide.dev
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)