quic bidistream benchmark shows bidistream would reduce send and recv packet rate about quic-go HOT 3 CLOSED

Streams don’t reduce their send rate, unless blocked by flow control. The connection might reduce the send rate due to congestion control.

I don’t know what you’re benchmarking here, but having thousands of connections between the same nodes doesn’t seem to be very useful. I’m happy to engage with actual issues, but “I created an unrealistic setup and things are slow maybe” is not one of them.

from quic-go.

Streams don’t reduce their send rate, unless blocked by flow control. The connection might reduce the send rate due to congestion control.

I don’t know what you’re benchmarking here, but having thousands of connections between the same nodes doesn’t seem to be very useful. I’m happy to engage with actual issues, but “I created an unrealistic setup and things are slow maybe” is not one of them.

My Purpose was to test the ability of a quic server to handle packets. Performance is expected to vary with the number of clients because three go routines are created for each quic long connection, and memory consumption is expected to vary, especially we use quicserver as long term connection. Is there some benchmark result or suggestion?

from quic-go.

Streams don’t reduce their send rate, unless blocked by flow control. The connection might reduce the send rate due to congestion control.

I don’t know what you’re benchmarking here, but having thousands of connections between the same nodes doesn’t seem to be very useful. I’m happy to engage with actual issues, but “I created an unrealistic setup and things are slow maybe” is not one of them.

We need to ensure the impact of a large number of quic long connections on system performance。

from quic-go.