DEV Community

Vikas Singhal
Vikas Singhal

Posted on

I Replaced Datadog With a 10MB Monitoring Tool (Here's What Happened)

#selfhosted #monitoring #opensource #devops

I have 4 servers. Nothing enterprise-grade - a couple of app servers, a database box, and a staging machine. Datadog was costing me $60/month to show me CPU graphs and disk usage.

That felt wrong.

So I tested alternatives. Grafana + Prometheus was the obvious choice, but the stack itself uses 500MB+ of RAM. On a $4/mo VPS with 2GB total, that's 25% of my resources just for monitoring.

Then I found Beszel. Its agent uses less than 10MB of RAM. The entire monitoring stack - hub and agents across 4 servers - runs on a $3/mo server and barely touches 50MB.

Here's what happened when I switched.

What Beszel Actually Monitors

Beszel tracks the basics well:

  • CPU usage and load average
  • Memory and swap usage
  • Disk usage and I/O
  • Network throughput
  • Docker container metrics (CPU/memory per container)

It does NOT do:

  • Application Performance Monitoring (APM)
  • Distributed tracing
  • Log aggregation
  • Custom metrics/dashboards beyond server basics
  • Alerting with complex conditions (basic alerts only)

If you need APM and tracing, Beszel isn't for you. If you need "is my server healthy?" at a glance, it's all you need.

The Setup (5 Minutes)

Hub (runs on one server - I used a $3/mo InstaPods server):

The hub is the web dashboard that collects and displays metrics. One-click deploy gives you a running instance with HTTPS in about 60 seconds.

Agent (runs on each monitored server):

curl -sL https://raw.githubusercontent.com/henrygd/beszel/main/supplemental/scripts/install-agent.sh -o install-agent.sh
chmod +x install-agent.sh
./install-agent.sh
Enter fullscreen mode Exit fullscreen mode

Enter your hub URL when prompted. The agent starts reporting metrics immediately.

Total setup time across 4 servers: about 5 minutes. Compare that to Prometheus + Grafana (30-60 minutes for the stack, then configuring dashboards, then setting up node_exporter on each server).

Resource Usage Comparison

I measured actual resource usage across monitoring tools on identical VPS instances (2 vCPU, 4GB RAM):

Tool Agent RAM Hub/Server RAM Total (4 agents + hub)
Beszel <10MB ~50MB ~90MB
Netdata 200-500MB N/A (each instance is standalone) 800-2000MB
Prometheus + node_exporter 50-100MB (exporter) 300-500MB (Prometheus) 500-900MB
Grafana (if you add it) N/A 200-400MB adds 200-400MB on top
Datadog agent 200-500MB N/A (cloud) 800-2000MB

Beszel's total footprint across my entire infrastructure (hub + 4 agents) is less than what a single Netdata or Datadog agent uses on one server.

What I Lost

Switching from Datadog means losing:

  • APM traces - I can't see request latency broken down by endpoint. For debugging slow API calls, I now SSH in and check logs manually.
  • Log search - Datadog's log search was convenient. I now use grep and journalctl on the servers directly.
  • Custom dashboards - Datadog lets you build anything. Beszel's dashboard is fixed (server metrics only).
  • Integrations - Datadog connects to everything (AWS, Kubernetes, databases). Beszel monitors the server, not specific services.
  • Alerts - Datadog's alerting is sophisticated (anomaly detection, composite conditions). Beszel has basic threshold alerts.

For my 4-server setup, none of these losses mattered. I wasn't using APM. My logs are small enough to grep. My dashboard needs are basic.

What I Gained

  • $57/month savings - $60/month (Datadog) vs $3/month (Beszel hub hosting). That's $684/year.
  • No data leaving my infrastructure - monitoring data stays on my servers. No third-party data processor.
  • Lower server overhead - freeing up 200-500MB per server means my apps have more headroom.
  • Simpler mental model - one dashboard, server metrics, done. No 47-tab Datadog interface.
  • No per-host pricing anxiety - adding a 5th server to Beszel costs $0 extra. Adding it to Datadog costs $15/month.

When Beszel Is Wrong

Don't switch to Beszel if:

  • You're running 50+ servers (you need the scalability of Prometheus/Grafana or a cloud solution)
  • You need APM with distributed tracing (use Datadog, New Relic, or self-hosted Jaeger)
  • You need log aggregation (use Loki, Elasticsearch, or a cloud logging service)
  • Your compliance requires specific monitoring certifications
  • You have a team of SREs who need advanced dashboards and runbooks

When Beszel Is Right

Switch to Beszel if:

  • You're monitoring 1-20 servers
  • You need CPU, memory, disk, network metrics
  • You want Docker container monitoring
  • Budget matters (your credit card, not the company's)
  • You value simplicity over features
  • You want to self-host your monitoring data

My Setup Now

  • Beszel hub: $3/mo managed server (InstaPods, one-click deploy)
  • 4 agents: installed in ~1 minute each, <10MB RAM each
  • Total cost: $3/month
  • What I check daily: CPU load, memory %, disk space, network throughput
  • Time saved vs Datadog: I spend less time in the monitoring UI because there's less to get distracted by

The honest answer: Beszel gives me 80% of what I used Datadog for at 5% of the cost. The other 20% (APM, logs, custom dashboards) I handle with SSH and grep when I need them, which is rarely.

What monitoring tool are you running? Curious if anyone else has downsized from Datadog/New Relic to something lighter - and whether you regret it.

Top comments (0)