Skip to content

Slow or Intermittent Connections

Use this guide after Module 14: Troubleshooting Method when a connection succeeds sometimes but performs poorly or fails unpredictably.

A single successful request proves only that one exchange worked. Intermittent problems require repeated observations, timestamps, and comparisons with a known working path.

Before collecting data, ask:

  • Is the problem constant, periodic, or random?
  • Does it affect one client, one application, or many users?
  • Does it happen on Wi-Fi, Ethernet, a virtual private network (VPN), or all connections?
  • Is it tied to a location, time of day, large transfer, or application action?
  • Did latency, packet loss, or throughput change from a known baseline?

Record the time and timezone of each occurrence. This lets an administrator compare the client symptoms with interface, firewall, access-point, server, and application logs.

Run a bounded series of tests instead of leaving ping running indefinitely.

Windows:

ping -n 20 <default-gateway-address>
ping -n 20 <known-remote-address>

Linux or macOS:

ping -c 20 <default-gateway-address>
ping -c 20 <known-remote-address>

Compare the results:

Pattern Where to look next
Loss or large response-time changes begin at the gateway Local Wi-Fi or Ethernet, switching, the client, or gateway load
Gateway results are stable but remote results are poor Routing beyond the local network, an upstream provider, or the remote path
Both tests are stable but the application is slow DNS, the destination port, Transport Layer Security (TLS), the application, or a backend dependency
Only one client is affected That client’s interface, driver, VPN, security software, configuration, or physical location

These patterns narrow the investigation but do not prove the cause. Internet Control Message Protocol (ICMP) traffic may be handled differently from application traffic.

If practical, compare the same client and application over Ethernet and Wi-Fi.

  • Poor Wi-Fi with stable Ethernet points toward signal strength, interference, channel use, roaming, or the wireless adapter.
  • Poor results on both connections point elsewhere, although both paths may still share the same gateway or internet connection.
  • A problem limited to one physical location may involve coverage or interference rather than the application.

Do not make several wireless changes at once. Record the original configuration and compare one controlled change at a time.

Ping does not measure application health. Repeat the exact operation that users report as slow, and compare it with a simple test for that application.

For a quick website test on Windows:

curl.exe -I https://example.com

On Linux or macOS:

curl -I https://example.com

A response confirms that the client resolved the name, connected to the server, negotiated TLS, and received an HTTP response. An error may instead identify a name-resolution, connection, or certificate problem. Replace the example address with the affected website when troubleshooting.

If the application uses a different protocol, use a client or health check that speaks that protocol.

If the earlier comparisons do not locate the delay:

  • Run tracert or traceroute and look for a consistent change, not one silent router.
  • Compare an affected client with a working client at the same time.
  • Check client, network-device, server, and application logs for the same timestamp.
  • Use Reading a Packet Capture when you have a specific question about loss, retransmissions, connection setup, or response timing.
  • Intermittent problems require repeated, timestamped observations rather than one successful test.
  • Comparing the gateway, a remote destination, and the actual application helps locate where performance changes begin.
  • Compare an affected client with a working client instead of drawing conclusions from one isolated result.