LinkLatencyTester

中文说明

Rust CLI utility for measuring link latency over TCP and UDP. It provides a symmetrical server/client pair with echo semantics, high-resolution RTT tracking, jitter/loss stats, concurrent TCP+UDP testing, and colorful tabular reports designed for real-world network troubleshooting.

Features

• Dual-protocol server – one process listens on both TCP & UDP (default port 9000), sharing the same Ctrl+C shutdown signal.
• Client echo tests – TCP payloads are length-prefixed, UDP payloads contain sequence & magic headers for validation.
• Accurate RTT & jitter – timestamps captured in milliseconds (f64) with jitter computed over deltas.
• Loss & timeout tracking – number of sent/received packets, loss %, and timed-out exchanges.
• Concurrent TCP+UDP measurements – --proto both (default) runs both protocols in parallel and prints a unified colored table.
• Non-blocking server logging – application logs are streamed through an async logger thread to avoid affecting RTT.
• Progress display – live progress line (TCP x/y | UDP x/y) while concurrent tests run.

Build

# Debug build
cargo build

# Release build
cargo build --release

Binaries are emitted into target/debug/ and target/release/ respectively.

Usage

Server

LinkLatencyTester.exe server [--proto both|tcp|udp] [--bind 0.0.0.0] \
    [--tcp-port 9000] [--udp-port 9000] [--echo true]

• --proto both (default) starts TCP & UDP listeners simultaneously.
• --echo true/false controls whether payloads are echoed back. Echo drives RTT measurement.

Example:

LinkLatencyTester.exe server --proto both --bind 0.0.0.0 --tcp-port 9000 --udp-port 9000

Client

LinkLatencyTester.exe client --host <server_ip> \
    [--proto both|tcp|udp] [--tcp-port 9000] [--udp-port 9000] \
    [--size 16] [--count 20] [--interval-ms 200] [--timeout-ms 1000]

• --proto both (default) runs TCP+UDP concurrently and prints one summary table.
• Single-protocol modes retain verbose per-packet logging (send/recv timestamps, RTT, len, errors).
• Progress line (e.g. TCP 7/20 | UDP 5/20) refreshes until both protocols finish.
• --plain forces ASCII borders + no colors/icons (for limited terminals) while keeping the default styling otherwise.

Example:

LinkLatencyTester.exe client --host 203.0.113.10 --proto both --count 30 --size 32

Traceroute hop listing

After the latency table, the client can optionally print a hop-by-hop list using the built-in tracert integration (default: enabled).

• --trace=true|false turns hop printing on or off.
• --trace-max-hops, --trace-wait-ms, --trace-send-rate-ms, and --trace-timeout-s map directly to the tracer settings.
• Output respects --plain, so hop rows stay ASCII-friendly when needed.

Windows requirements: tracert uses raw sockets. Run the CLI from an elevated PowerShell/Command Prompt and allow ICMP "Time-to-live Exceeded" / "Destination Unreachable" in Windows Firewall. You can temporarily enable this with:

netsh advfirewall firewall add rule name="All ICMP v4" dir=in action=allow protocol=icmpv4:any,any
netsh advfirewall firewall add rule name="All ICMP v6" dir=in action=allow protocol=icmpv6:any,any

Example command that runs latency tests plus traceroute:

LinkLatencyTester.exe client --host 203.0.113.10 --proto both --count 20 --trace --trace-max-hops 30

Sample Output

TCP 20/20 | UDP 20/20
Target: tcp=203.0.113.10:9000 udp=203.0.113.10:9000
┌───────────┬─────┬─────┬─────┬──────┬────────┬────────┬─────────┬────────┬────┐
│ PROTO     ┆ SENT┆ RECV┆ LOST┆ LOSS%│ MIN(ms)│ AVG(ms)│ MAX(ms) │ JIT(ms)│ TO │
╞═══════════╪═════╪═════╪═════╪══════╪════════╪════════╪═════════╪════════╪════╡
│ TCP       ┆ 20  ┆ 19  ┆ 1   ┆ 5.00 │ 63.286 │ 678.392│ 1122.932│ 126.247│ 1  │
├───────────┼─────┼─────┼─────┼──────┼────────┼────────┼─────────┼────────┼────┤
│ UDP       ┆ 20  ┆ 16  ┆ 4   ┆ 20.00│ 107.075│ 114.168│ 119.912 │ 3.076  │ 4  │
└───────────┴─────┴─────┴─────┴──────┴────────┴────────┴─────────┴────────┴────┘

Colors highlight good (green), warning (yellow), or poor (red) performance thresholds.

Need compatibility mode? Append --plain to the client command above to render the ASCII table without colors or box-drawing glyphs.

Traceroute hop list (excerpt):

Trace (tracert): 203.0.113.10 (max_hops=30 wait_ms=1500 send_rate_ms=0 timeout_s=30)
status: Done
 1)          192.168.1.1 hop=1 ttl=64 rtt=2.7 ms (Δn/a) default-gateway
 2)            10.10.0.2 hop=2 ttl=63 rtt=4.1 ms (Δ+1.4 ms) relay
 3)         125.68.140.1 hop=3 ttl=253 rtt=9.3 ms (Δ+5.2 ms) relay
 4)      182.134.236.157 hop=4 ttl=252 rtt=15.0 ms (Δ+5.7 ms) relay
 5)         202.97.29.33 hop=5 ttl=250 rtt=42.1 ms (Δ+27.1 ms) relay
 6)        121.189.3.105 hop=6 ttl=246 rtt=88.4 ms (Δ+46.3 ms) relay
 7)       203.0.113.10 hop=7 ttl=242 rtt=95.0 ms (Δ+6.6 ms) destination
trace duration: 8.36s

Parameters

Option	Default	Description
--proto	both	Choose tcp, udp, or both for dual-mode.
--bind (server)	0.0.0.0	Address to bind for TCP/UDP listeners.
--tcp-port / --udp-port	9000	Ports for each protocol.
--echo (server)	true	Whether to echo payloads back.
--host (client)	required	Target server hostname/IP.
--size	16	Payload size in bytes (>= 8 reserved for header).
--count	20	Number of packets per protocol.
--interval-ms	200	Delay between packets.
--timeout-ms	1000	Socket read/write timeout per packet.

Implementation Notes

• TCP payload framing (u32 length + payload) prevents sticky packets and preserves 1:1 RTT mapping.
• UDP payload embeds sequence+magic to validate responses.
• Server log output uses a background thread (mpsc channel) to avoid affecting response latency.
• Ctrl+C handler flips an AtomicBool, allowing graceful shutdown of both protocols.

Feel free to clone, tweak defaults, or extend the tool for your latency testing workflows.