High-Bandwidth BGP Tunnel Buyer's Guide: Calculating Oversubscription Ratios & Single-Thread Throttling

A technical buyer's guide to high-bandwidth budget BGP tunnel proxies, explaining oversubscription math, TCP congestion window collapse, and single-thread iperf3 diagnostics.


Benchmark Overview & Key Finding (TL;DR): Budget proxy providers marketing "¥15/month for 2,000GB" commonly operate with oversubscription ratios exceeding 40:1 to 60:1. While synthetic multi-thread speedtests present illusory 500Mbps results, real-world single-thread throughput for 4K streaming and software downloads collapses to under 5 Mbps during evening peak congestion. Auditing actual capacity requires measuring single-thread TCP throughput.


1. The Economics of Extreme Oversubscription

A 1Gbps domestic BGP relay costs ~¥16,000/month. At ¥15/subscriber, 1,200+ users are required to break even. This allocates a theoretical concurrency floor of just 0.83 Mbps per user.


2. Multi-Thread Speedtest Illusions vs Single-Thread Reality

Speedtest benchmarks spawn 16–32 parallel sockets, aggregating fragmented bandwidth. In contrast, YouTube video chunks and Git transfers rely on single TCP streams, which suffer immediate window degradation upon encountering packet loss.


3. Ecosystem Linking Matrix

RECOMMENDED SAFE ALTERNATIVE

Guangsu Cloud: Dual 2.5Gbps IEPL Dedicated Lines

Tested with 786.4 Mbps single-thread throughput and zero packet loss during peak hours. Featuring the Snooker 147 permanent 247GB package (pay once, never expires) and lightweight annual plan at ¥99/year. Use exclusive code AMM at checkout.