Skip to content

Latest commit

 

History

History
67 lines (50 loc) · 3.01 KB

README.md

File metadata and controls

67 lines (50 loc) · 3.01 KB
Raw

echoping

A golang network utility using TCP/UDP-Datagram/QUIC messages to detect network stability.

Why to write this network utility

It's difficult to detect network problems in real life.

echoping uses TCP and UDP(Datagram or QUIC) protocols instead of ICMP used by traditional ping, and send messages more frequently.

echoping works more like a real network application, and reports network problems as the real one.

How it works

  1. Start a echoping server by ./echoping -listen IP:PORT
  2. Start echoping clients to use the echoping server: ./echoping -connect SERVER_IP:SERVER_PORT
  3. Every echoping client sends TCP/UDP-Datagram/QUIC requests to the server
  4. The server sends every request as a response back to the client
  5. The client collects the responses and calculates the packet loss ratio (UDP only) and round-trip time
  6. If any network error occurs, the client will retry to connect and send requests forever.
  7. Every connection (ping session) has a unique session id which is generated by the connection time, it helps to distinguish different client connections, eg: if a network error occurs and a client re-connects, then there will be a new session id in logs.

⚠️ The server and client versions must match, otherwise the protocol may not work ⚠️

Usage:

  -connect string
    	Connect to 'tcp://ip:port/,udp://ip:port/,quic://ip:port/' (can be repeated, use comma as delimiter), or use 'ip:port' for all TCP/UDP/QUIC
  -listen string
    	Listen both TCP and UDP on ip:port (UDP also works for QUIC)
  -listen-tcp string
    	Listen TCP on ip:port
  -listen-udp string
    	Listen UDP on ip:port (UDP also works for QUIC)
  -loss-ratio string
    	The simulated UDP loss ratio on client side (must be used with "-connect"). "0.1"" means 10% packet loss, "0.1,0.2" means 0.1 for sending and 0.2 for receiving
  -payload-size int
    	The payload size of ping request (default 1300)
  -ping-interval duration
    	The interval between ping requests sent by client (default 20ms)

Server output:

2020/12/26 01:24:54 server stat udp:127.0.0.1:52032 (20201226-012449.458182): pps=10.0, tmperr=0
2020/12/26 01:24:54 server stat tcp:127.0.0.1:56524 (20201226-012449.458231): pps=10.0, tmperr=0

Client output:

2020/12/26 01:24:54 client stat udp:127.0.0.1:12345 (20201226-012449.458182): pps=11.0, recv=0.01MB/s, loss=0.0%, round-trip time (ms): avg=0.4, min=0.2, max=0.6, stddev=0.1
2020/12/26 01:24:54 client stat tcp:127.0.0.1:12345 (20201226-012449.458231): pps=11.0, recv=0.01MB/s, loss=0.0%, round-trip time (ms): avg=0.3, min=0.2, max=0.4, stddev=0.1
  • "20201226-012449.458182" is the ping session id, which is generated by the time of the client starts the ping session.
  • pps means "ping per second"
  • The loss of TCP and QUIC is always 0% (of course)

Develop and debug locally

./echoping -listen :12345 -connect 127.0.0.1:12345

Simulate delayed and dropped packets on Linux: https://stackoverflow.com/questions/614795/simulate-delayed-and-dropped-packets-on-linux