Characterizing the workload of a netflix streaming video server
Bibliographic record
Abstract
In this paper we characterize the workload of a Netflix streaming video web server. Netflix is a widely popular subscription service with over 81 million global subscribers [24]. The service streams professionally produced TV shows and movies over the Internet to an extremely diverse and representative set of playback devices over broadband, DSL, WiFi and cellular connections. Characterizing this type of workload is an important step to understanding and optimizing the performance of the servers used to support the growing number of streaming video services. We focus on the HTTP requests observed at the server from Netflix client devices by analyzing anonymized log files obtained from a server containing a portion of the Netflix catalog. We introduce the notion of chains of sequential requests to represent the spatial locality of the workload and find that despite servicing clients that adapt to changes in network and server conditions, and despite the fact that the majority of chains are short (60% are no longer than 1 MB), the vast majority of the bytes requested are sequential. We also observe that during a viewing session, client devices behave in recognizable patterns. We characterize sessions using transient, stable and inactive phases. We find that playback sessions are surprisingly stable; across all sessions 5% of the total session time is spent in transient phases, 79% in stable phases and 16% in inactive phases, and the average duration of a stable phase is 8.5 minutes. Finally we analyze the chains to evaluate different prefetch algorithms and show that by exploiting knowledge about workload characteristics, the workload can be serviced with 13% lower hard drive utilization or 30% less system memory compared to a prefetch algorithm that makes no use of workload characteristics.
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 | 0.005 |
| 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 | 0.002 |
| Open science | 0.001 | 0.000 |
| Research integrity | 0.001 | 0.000 |
| 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".