Squash: low latency multi-path video streaming using multi-bitrate encoding
Bibliographic record
Abstract
The demand for low latency video streaming has dramatically increased as live video streaming applications, such as Twitch and Youtube Live, are becoming more popular. According to the 2021 Bitmovin video developer report, the biggest challenge that video developers are experiencing today is providing low latency video streaming. One of the most common on-site live streaming methods is using a wireless LTE network. There have been many approaches for characterizing wireless links and accurately measuring available bandwidth to provide low latency streaming over a wireless LTE network link. However, even with fine-grained bandwidth estimation, video streaming on a single LTE link is still susceptible to unexpected network delay from a sudden drop in available bandwidth or temporal disconnection. \n \nPeople can utilize multiple wireless LTE links to overcome the limitations of using a single LTE link for low latency video streaming. Using multiple links can enhance video quality through increased bandwidth and resilience. However, multi-homed low latency video streaming protocols may achieve lower video quality than single-homed protocols when a frame is split and sent over more than one link. Suppose one of the links becomes congested or gets disconnected. In that case, the part of the frame sent on stable links must wait until the packets sent on the problematic link are re-transmitted through another link. Re-transmission requires at least one extra round trip time. A video player may skip the late frame or serve only the received part of the frame due to the re-transmission delay. Ferlin et al. suggest using Forward Error Correction (FEC) on Multipath TCP (MPTCP) to reduce re-transmission delay. However, FEC is not helpful in the event of a significant bandwidth drop. If the sender does not use sufficient redundancy to handle a significant bandwidth drop, the receiver will not receive enough blocks to decode the video data. FEC requires using a large portion of the network bandwidth for redundancy to handle significant bandwidth drops even when the links are stable. \n \nIn this thesis, I present Squash, a low latency video transport protocol that encodes each frame at multiple bitrates and sends them across different links to minimize video stream disruption in the event of unexpected bandwidth drops. The encoder encodes a frame into multiple different bitrates, which are high-bitrate and low-bitrate. When a high- bitrate frame cannot arrive on time due to congestion from an unexpected drop in available bandwidth, the low-bitrate frame is used to replace the missing frame. This is because the low-bitrate frame is smaller and is sent on the links that are disjoint from those used by the high-bitrate frame. To the best of my knowledge, Squash is the first architecture that uses multi-bitrate frames to increase resilience against unexpected bandwidth drops in low latency video streaming over multiple wireless LTE links. In emulated wireless LTE network environment using Mahimahi network traces, the average SSIM of the video streamed on Squash is 13 – 58% higher than that streamed on the baseline protocol, which is designed in the same manner as Squash except that it employs single-frame encoding.
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 distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.001 | 0.000 |
| Scholarly communication | 0.000 | 0.001 |
| Open science | 0.001 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 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 teacher head, 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".