To the QNAP Development and Network Engineering Team,
I am writing to formally request the implementation of Google’s BBR (Bottleneck Bandwidth and Round-trip propagation time) as the default TCP congestion control algorithm in QTS/QuTS hero, or at the very least, the addition of a user-selectable toggle in the “Network & Virtual Switch” GUI.
Currently, QTS defaults to TCP CUBIC. While CUBIC has been the historical standard, it severely underperforms in modern LFNs (Long Fat Networks) and high-latency WAN/VPN scenarios due to its loss-based nature.
My Infrastructure & Use Case:
-
NAS: QNAP TS-473A (16GB RAM, QTS 5.2.9)
-
Routing: MikroTik RB5009 routers on both ends
-
Topology: Site-to-Site WireGuard VPN over a 1 Gbps FTTH connection (routing through international ISP nodes, specifically TIM/O2, with moderate latency and occasional minimal packet drop).
-
Workload: High-bitrate media streaming (Plex Direct Play) and large file transfers over the WireGuard tunnel.
The Problem with CUBIC: During standard operations, my WireGuard tunnel throughput was drastically bottlenecked to around 39 Mbps. Extensive iperf3 testing revealed that CUBIC was aggressively halving the congestion window due to minor, unavoidable packet loss along the international transit nodes. The loss-based algorithm choked the available bandwidth, causing thousands of unnecessary retransmissions and making high-bitrate streaming impossible without transcoding (which the Ryzen V1500B must handle entirely via CPU, lacking an iGPU).
The Solution & Empirical Results (BBR): To diagnose the issue, I manually altered the kernel parameters via SSH, applying: sysctl -w net.core.default_qdisc=fq sysctl -w net.ipv4.tcp_congestion_control=bbr (To make it persistent, I had to create a custom script triggered via crontab, as QTS resets sysctl.conf upon reboot).
The results were immediate and staggering. By switching from a loss-based to a latency-based algorithm:
-
Throughput increased by 10x: from ~39 Mbps to ~380/400 Mbps over the exact same encrypted WireGuard tunnel.
-
Stability: Packet delivery smoothed out entirely, eliminating buffering for 4K Direct Play media.
The Request: Modern Linux distributions, Windows 11, and major cloud providers have already migrated to BBR for its superior handling of real-world internet topologies. QTS is fundamentally a robust Linux system, and the kernel already supports the bbr module.
I kindly ask the engineering team to consider:
-
Migrating the default TCP congestion control from CUBIC to BBR natively.
-
Alternatively, exposing a simple dropdown menu in the Network & Virtual Switch interface to allow advanced users to switch between CUBIC and BBR without resorting to CLI crontab workarounds that might be overwritten by firmware updates.
Thank you for your time and for continuing to build excellent hardware. I look forward to your feedback on this architectural improvement.