Aquila measures what your line actually delivers: download, upload, latency –
and the bufferbloat that explains why everything stalls the moment an upload
is running.
A run takes about 25 seconds. One dial, two numbers, one line of latency.
WHAT IS MEASURED
• Download and upload in megabits per second
• Latency of the idle line – median of seven probes
• Latency under load: the same measurement while download and upload run
• Bufferbloat: the difference between the two. If your latency rises from
20 to 300 milliseconds under load, that is why video calls stutter –
and it is not a question of bandwidth
MEASURED HONESTLY
• Six streams down, four up – a single connection is limited by its window,
not by your line
• A separate connection per stream, otherwise you measure HTTP/2 multiplexing
• The first two seconds do not count: they show the ramp-up
• Random data on upload – zeros would give dream results on compressing links
• Latency probes on their own connection, otherwise you measure your own
queue instead of the line's bufferbloat
• Median instead of mean: one outlier must not shift the verdict
HISTORY
• Every measurement is kept with date, place and network – Wi-Fi name,
mobile carrier or LAN
• So you can tell whether it is the line or the location
• The history stays on your device
MADE TO LOOK AT
• Eggshell paper, a black engraved instrument, one coloured glow – no neon,
no data sheet
• Freely chosen colour across the full wheel, from red through yellow and
green to violet
• Black-and-white mode and an inverted night version
• 34 languages
PRIVATE
• No account, no ads, no analytics or tracking
• The measurement history never leaves your device
• Location is used only to name the place in your history – it is never
transmitted. Without permission Aquila measures just fine
Measurements run against Cloudflare's public measurement endpoint.
Measure speed, latency and bufferbloat. No account, no ads, no tracking.