Debugging a “Slow VPS” That Wasn’t Slow

Public
/system-design

My VPS suddenly felt extremely slow:

  • pages taking 3–5 seconds
  • TLS handshake hanging
  • random spikes in latency

First instinct: server issue.

Reality: network path instability.

This is how I debugged it step by step.

1. Measure Real User Experience with curl

Instead of guessing, start with actual HTTP timing.

  
  
curl -o /dev/null -s -w "\
DNS: %{time_namelookup}s\n\
Connect: %{time_connect}s\n\
TLS: %{time_appconnect}s\n\
TTFB: %{time_starttransfer}s\n\
Total: %{time_total}s\n" \
https://your-domain.com

What each metric means

  • DNS (time_namelookup) → time to resolve domain → IP

  • Connect (time_connect) → TCP handshake (network latency baseline)

  • TLS (time_appconnect) → SSL/TLS negotiation (very sensitive to packet loss)

  • TTFB (time_starttransfer) → time until first byte from server

  • Total (time_total) → full request time

Example (bad case)

  
  
DNS: 0.05s
Connect: 0.30s
TLS: 2.29s
TTFB: 3.10s
Total: 5.01s

This is NOT normal.

2. Isolate Network vs Server

Force IPv4 (avoid dual-stack ambiguity):

  
  
curl -4 -o /dev/null -s -w "Connect:%{time_connect}s TLS:%{time_appconnect}s TTFB:%{time_starttransfer}s Total:%{time_total}s\n" https://your-domain.com

If this is stable, the issue is not your backend.

3. Check Your Server Internally

From inside the VPS:

  
  
curl -o /dev/null -s -w "TTFB: %{time_starttransfer}s\nTotal: %{time_total}s\n" http://localhost

If this is ~1ms:

→ your app is NOT the problem.

4. Measure Raw Network Latency with ping

  
  
ping -c 100 your-server-ip

Look for:

  • packet loss %
  • average latency
  • jitter (min/max spread)

Example (bad)

  
  
13% packet loss

Even 1–2% can hurt.
10%+ will destroy performance.

5. Understand the Path with traceroute

Before using MTR, a quick traceroute gives you a snapshot of the route.

  
  
traceroute -n your-server-ip

Why -n matters

  • Skips DNS resolution
  • Faster
  • Cleaner output
  • Avoids misleading delays caused by reverse DNS

What traceroute shows

Each line = one hop (router)

Example:

  
  
1  192.168.1.1
2  80.10.x.x
3  orange.net
4  cogentco.com
5  ...

Important notes

  • * * * = router not replying (NOT necessarily packet loss)
  • Latency jumps = geographic distance or congestion
  • Paths can be asymmetric (forward ≠ return path)

Example insight

In my case:

  
  
Paris → Cogent → Washington → Atlanta → Texas → Phoenix

→ My VPS was in the US, not Europe.

6. Trace the Network Path with mtr

  
  
mtr -4 -rw your-server-ip

Important flags

  • -4 → force IPv4
  • -r → report mode (no interactive UI)
  • -w → wide output (no truncation)
  • -c 100 → number of packets (recommended for accuracy)

Example:

  
  
mtr -4 -rw -c 100 your-server-ip

How to read MTR

Each row = one hop (router)

Columns:

  • Loss% → packet loss at that hop
  • Avg → average latency
  • Best/Wrst → jitter range

Critical rule

Intermediate hop loss often lies.

Routers may:

  • rate-limit ICMP
  • ignore probes

What matters:

  • Final hop loss
  • End-to-end ping results

Example (real issue)

  
  
Final hop: 4% loss
Avg: 204ms
Best: 146ms
Worst: 236ms

This means:

  • real packet loss
  • high jitter
  • unstable path

7. Understand the Root Cause

In my case:

  • VPS was in the US
  • I was in Europe
  • traffic path:
  
  
France → Cogent → US backbone → Phoenix → VPS

Issues observed:

  • packet loss spikes (0% → 20% → 0%)
  • unstable routing
  • TLS delays (because retransmissions)

8. Why TLS Explodes Under Packet Loss

TLS requires multiple round trips.

With packet loss:

  • packets are retransmitted
  • handshake stalls
  • latency multiplies

That’s why:

  
  
Ping: 250ms
→ feels like seconds in real requests

9. Final Outcome

After some time:

  • packet loss dropped to 0%
  • curl stabilized:
  
  
Connect: 0.15s
TLS: 0.30s
TTFB: 0.45s
Total: 0.60s

Conclusion:

→ transient network congestion / routing issue

10. Key Takeaways

  • Always measure before guessing
  • Separate:
    • server performance
    • network latency
    • packet loss
  • Use:
    • curl → user experience
    • ping → packet loss
    • mtr → path analysis
  • Packet loss hurts more than latency
  • Geography still matters

11. One-Liner Summary

My VPS wasn’t slow. The Internet path to it was.