Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

fix(ndt7-protocol): acknowledge the importance of TCP_INFO #388

Open
wants to merge 2 commits into
base: main
Choose a base branch
from
Open

fix(ndt7-protocol): acknowledge the importance of TCP_INFO #388

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
706aedb
fix(ndt7-protocol): acknowledge the importance of TCP_INFO
bassosimone Jun 7, 2023
b4b8ae2
Update ndt7-protocol.md
bassosimone Jun 7, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
18 changes: 8 additions & 10 deletions spec/ndt7-protocol.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -7,7 +7,7 @@ protocol](https://github.com/ndt-project/ndt). Ndt7 is based on
WebSocket and TLS, and takes advantage of TCP BBR, where this
flavour of TCP is available.

This is version v0.9.1 of the ndt7 specification.
This is version v0.10.0 of the ndt7 specification.

## Design choices

Expand Down Expand Up @@ -35,11 +35,13 @@ a measurement of your last mile speed. Rather it is a measurement
of what performance is possible with your device, your current internet
connection (landline, Wi-Fi, 4G, etc.), the characteristics of
your ISP and possibly of other ISPs in the middle, and the server
being used. The main metric measured by ndt7 is the goodput. That is,
the speed measured at application level, without including the
overheads of WebSockets, TLS, TCP/IP, and link layer headers. But we
also provide kernel-level information from `TCP_INFO` where available. For
all these reasons we say that ndt7 performs application-level measurements.
being used. The original design of ndt7 was meant to support client
and server only measuring the goodput (i.e., the speed measured at
the application level, without including the overhead of WebSocket,
TLS, TCP/IP, and link layer headers). However, the M-Lab deployment of
ndt7 includes goodput measurements as well as `TCP_INFO` measurements
collected by the server. This implementation choice allows for
maximally simpler clients and pushes the complexity to the server.

The presence of network issues (e.g. interference or congestion) should
cause ndt7 to yield worse measurement results, relative to the expected speed
Expand All @@ -50,10 +52,6 @@ Extra information obtained using `TCP_INFO` should help an expert
reading the results of a ndt7 experiment to better understand what could
be the root cause of such performance issues.

To ensure that clients continue to work, we make the design choice that
clients should be maximally simple, and that all complexity should
implemented on the server side.

Ndt7 should consume few resources. The maximum runtime of a test should
be ten seconds, but the server should be able to determine if the performance
if performance has stabilized in less than ten seconds and end the test early.
Expand Down