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

test(transport): assert maximum bandwidth on gbit link #2203

Open
wants to merge 1 commit into
base: main
Choose a base branch
from
Open

test(transport): assert maximum bandwidth on gbit link #2203

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
44 changes: 44 additions & 0 deletions neqo-transport/tests/network.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -196,3 +196,47 @@ fn transfer_fixed_seed() {
sim.seed_str("117f65d90ee5c1a7fb685f3af502c7730ba5d31866b758d98f5e3c2117cf9b86");
sim.run();
}

#[test]
#[allow(clippy::cast_precision_loss)]
fn gbit_bandwidth() {
const MIB: usize = 1024 * 1024;
const TRANSFER_AMOUNT: usize = 100 * MIB;
Copy link
Collaborator Author

@mxinden mxinden Oct 26, 2024 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Takes ~2s to run on my machine, 1s for up-, 1s for download. Using < 100 MIB doesn't give me consistent results.

Worth spending 2s of our unit-test runtime on this?

(Not to be confused with simulated time.)


for upload in [false, true] {
let sim = Simulator::new(
"gbit-bandwidth",
boxed![
ConnectionNode::default_client(if upload {
boxed![SendData::new(TRANSFER_AMOUNT)]
} else {
boxed![ReceiveData::new(TRANSFER_AMOUNT)]
}),
TailDrop::gbit_link(),
ConnectionNode::default_server(if upload {
boxed![ReceiveData::new(TRANSFER_AMOUNT)]
} else {
boxed![SendData::new(TRANSFER_AMOUNT)]
}),
TailDrop::gbit_link()
],
);

let simulated_time = sim.setup().run();
let bandwidth = TRANSFER_AMOUNT as f64 * 8.0 / simulated_time.as_secs_f64();

// Given Neqo's current static stream receive buffer of 1MiB, maximum
// bandwidth is below gbit link bandwidth.
//
// Tracked in https://github.com/mozilla/neqo/issues/733.
let maximum_bandwidth = MIB as f64 * 8.0 / 0.1; // bandwidth-delay-product / delay = bandwidth
let expected_utilization = 0.5;
Copy link
Collaborator Author

@mxinden mxinden Oct 26, 2024 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Running this test in a loop, Neqo is not able to utilize more than ~50% of the maximum bandwidth of 80 Mbit/s. Intuitively, even with packet loss, I would expect a congestion controller to be able to saturate more than 50% of a link.

45.95196838717057
49.31818797697269
45.826026606578566
49.53933723888579
45.997472274917506
49.31788343565587
45.58981370339061
49.31818797697269
40.48485804428038
49.451830153095756
40.5585726298109
49.31793477754305
40.55224164532084
54.71429614593755
45.57110592947965
49.45192167242847
40.53596208031323
54.79092558021697
45.95144831798395

I will look for mistakes in my math with a fresh mind tomorrow.

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Is this using Reno or Cubic? Pacing or not?


assert!(
maximum_bandwidth * expected_utilization < bandwidth,
"with upload {upload} expected to reach {expected_utilization} of maximum bandwidth ({} Mbit/s) but got {} Mbit/s",
maximum_bandwidth / MIB as f64,
bandwidth / MIB as f64,
);
}
}
4 changes: 3 additions & 1 deletion test-fixture/src/sim/mod.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -293,7 +293,8 @@ pub struct ReadySimulator {
}

impl ReadySimulator {
pub fn run(mut self) {
#[allow(clippy::must_use_candidate)]
pub fn run(mut self) -> Duration {
let real_start = Instant::now();
let end = self.sim.process_loop(self.start, self.now);
let sim_time = end - self.now;
Expand All @@ -303,5 +304,6 @@ impl ReadySimulator {
wall = real_start.elapsed(),
);
self.sim.print_summary();
sim_time
}
}
10 changes: 10 additions & 0 deletions test-fixture/src/sim/taildrop.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -90,6 +90,16 @@ impl TailDrop {
Self::new(200_000, 8_192, Duration::from_millis(50))
}

/// A tail drop queue on a 1 Gbps link with the default forward delay of
/// 50ms and a buffer equal to the bandwidth-delay-product.
#[must_use]
pub const fn gbit_link() -> Self {
let rate = 1_000_000_000 / 8;
let delay = Duration::from_millis(50);
let capacity = rate / 20; // rate * 0.05
Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Simply the bandwidth-delay-product. Happy for more realistic suggestions.

Self::new(rate, capacity, delay)
}

/// How "big" is this datagram, accounting for overheads.
/// This approximates by using the same overhead for storing in the queue
/// and for sending on the wire.
Expand Down
Loading