TURBOTEST: Learning When Less is Enough through Early Termination of Internet Speed Tests

TL;DR

This work tackles the large data overhead of Internet speed tests by introducing TurboTest, an external termination framework that stops ongoing tests early without sacrificing accuracy. It decouples throughput prediction (Stage 1: regression on partial measurements) from the termination decision (Stage 2: classification), and enriches the feature space with RTT, retransmissions, and congestion-window signals alongside throughput, all governed by a single tolerance parameter $\epsilon$. Evaluated on about 1 million M-Lab NDT tests, TurboTest consistently pushes the accuracy–savings frontier farther than existing heuristics (BBR, CIS), achieving up to ~95% data savings with acceptable relative error and approaching oracle-like performance. The results demonstrate that adaptive ML-based termination can deliver accurate, scalable, and deployable speed tests, significantly reducing measurement overhead in practice while maintaining reliability across diverse networks.

Abstract

Internet speed tests are indispensable for users, ISPs, and policymakers, but their static flooding-based design imposes growing costs: a single high-speed test can transfer hundreds of megabytes, and collectively, platforms like Ookla, M-Lab, and Fast.com generate petabytes of traffic each month. Reducing this burden requires deciding when a test can be stopped early without sacrificing accuracy. We frame this as an optimal stopping problem and show that existing heuristics-static thresholds, BBR pipe-full signals, or throughput stability rules from Fast.com and FastBTS-capture only a narrow portion of the achievable accuracy-savings trade-off. This paper introduces TURBOTEST, a systematic framework for speed test termination that sits atop existing platforms. The key idea is to decouple throughput prediction (Stage 1) from test termination (Stage 2): Stage 1 trains a regressor to estimate final throughput from partial measurements, while Stage 2 trains a classifier to decide when sufficient evidence has accumulated to stop. Leveraging richer transport-level features (RTT, retransmissions, congestion window) alongside throughput, TURBOTEST exposes a single tunable parameter for accuracy tolerance and includes a fallback mechanism for high-variability cases. Evaluation on 173,000 M-Lab NDT speed tests (2024-2025) shows that TURBOTEST achieves nearly 2-4x higher data savings than an approach based on BBR signals while reducing median error. These results demonstrate that adaptive ML-based termination can deliver accurate, efficient, and deployable speed tests at scale.

Figures (9)

• Figure 1: TurboTest workflow
• Figure 2: Distribution of tests across different speed tiers. The left bars show the fraction of total tests in each group, while the right bars show the fraction of total data transferred.
• Figure 3: Pareto frontiers of TurboTest, BBR, and CIS.
• Figure 4: Distribution of data transfer and relative errors across samples for parameters that satisfy the median error $<20\%$ constraint.
• Figure 5: Delta in data transfer between TurboTest and BBR across all speed-tier × RTT groupings. Point size indicates the magnitude of the difference in data transferred. Green denotes cases where TurboTest transfers less, while red denotes cases where BBR transfers less.