performance-utils

Measure the performance (latency) of your hosted software across all 7 layers of the OSI model, and see where you can optimize.

Two Python scripts, stdlib-only, work on Windows / macOS / Linux:

File	Purpose
network_layers_latency.py	Prints a plain-text 7-layer latency gauge in the terminal.
html_report.py	Runs the same measurements and writes a self-contained HTML report with an Ask Gemini panel.

Layers measured

Layer	OSI layer	What it measures
L1	Physical	Local NIC link speed (Wi-Fi / Ethernet)
L2	Data Link	Local MAC address and per-interface byte counters
L3	Network	ICMP ping + traceroute to the target host
L4	Transport	TCP connect latency to the target host:port
L5	Session	TLS handshake negotiation time
L6	Presentation	Negotiated TLS cipher + protocol version
L7	Application	HTTP request: DNS+TCP+TLS, TTFB, and total time

L5 and L6 share one measurement when the URL is https://; for plain http:// there is no TLS.

Requirements

• Python 3.8+ — no third-party packages, standard library only.
• A network connection to the target URL.
• Optional: traceroute on Linux/macOS, PowerShell Get-NetAdapter on Windows (for richer L1/L2 detail).

Quick start

# Default target (Google's 204 endpoint)
python network_layers_latency.py

# Plain-text measurement of your own URL
python network_layers_latency.py --url https://my-webapp.example.com

# Self-contained HTML report
python html_report.py --url https://my-webapp.example.com --html report.html

You can also set the target via TARGET_URL env var.

Options (both scripts)

Flag	Long	Default	Description
-u	--url	env / Google 204	Target URL to measure.
-n	--pings	4	Number of ICMP ping packets (L3).
-H	--hops	5	Max traceroute hops (L3).
-t	--timeout	5	Per-command / socket timeout (seconds).
-o	--html	stdout	html_report.py only: write report to this path.
	--model	gemini-3.6-flash	html_report.py only: model pre-filled in the Ask Gemini panel.

Ask Gemini

Every HTML report includes an Ask Gemini panel at the bottom. To use it:

1. Get a free API key at Google AI Studio.
2. Open the report in your browser, paste the key into the panel, edit the question if you want, and click Ask Gemini.

The key is stored only in this browser's localStorage and sent directly from your browser to Google's API — nothing passes through any other server. The model defaults to gemini-3.6-flash and can be changed per call.

Interpreting results

• L4 high, L3 low → server-side TCP accept queue / SYN handling.
• L5/L6 high → certificate chain, TLS version, client CPU. TLSv1.3 should be fast.
• L7 TTFB − (L4+L5+L6) large → application server is slow before sending any body (DB query, render, upstream call, etc.).

Notes

• L1/L2 are inherently local-machine measurements. L3–L7 measure the path to the target URL.
• The scripts are dependency-free so they run anywhere Python does.