Deep Learning-Based Fault Localization in Video Networks Using Only Client-Side QoE
Bibliographic record
Abstract
Maintaining a satisfactory customer Quality of Experience (QoE) is of vital importance for video service providers such as Netflix or Amazon Prime Video. Network faults degrade QoE and must therefore be detected, isolated, and fixed. However, this is difficult because each part of the end-to-end path belongs to a different autonomous system (AS) that is typically owned by a different entity, such as the video streaming provider, the internet service provider (ISP), and the client's local network operator. Although the video service provider (VSP) is usually blamed by the customer when there is poor QoE, the VSP does not have access to many parts of the network to localize the issue. In this paper, we show that with the aid of AI, it is possible for the VSP to localize the network fault without having access to the faulty part and using only QoE metrics. We collected a dataset from an actual video streaming testbed, where multiple videos are streamed from a video server through a simplified ISP network to a client network. Actual faults were generated in both the ISP and the client networks. Using only the QoE metrics measured at the client side, we use the deep learning methods of multi-layer perceptron (MLP) and long-short-term memory (LSTM) to detect and localize the fault with an accuracy of 93–97%, depending on the situation.Impact Statement—Technologically, our work impacts video/game streaming service providers such as Netflix, YouTube, Amazon Prime, Google Stadia, Sony PlayStation Now, Nvidia GeForce Now, and videoconferencing providers such as Zoom and Skype. Our work enables these providers to train similar AI systems that can localize network problems using only the video quality of experience (QoE) recorded by their client software. They can then take an appropriate action, such as rerouting traffic using Open Connect Appliances (OCA) if available, using another network provider if they have contracts with more than one, or informing the owner of the network segment with the fault, so they can fix the problem and maintain their customers’ QoE at a satisfactory level. Economically, our work can contribute to the market expansion of any video streaming solution because it will lead to better QoE, which is synonymous with more customers.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.003 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.001 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".