Web Experience in Mobile Networks: Lessons from Two Million Page Visits

Rajiullah, Mohammad; Lutu, Andra; Khatouni, Ali Safari; Fida, Mah-Rukh; Mellia, Marco; Brunström, Anna; Alay, Özgü; Alfredsson, Stefan; Mancuso, Vincenzo

doi:10.1145/3308558.3313606

Cited by 43 publications

(29 citation statements)

References 44 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Figure 1, rows (1) and (2) describe this testbed and its usage. Different from previous studies [1], [2], [15]- [17], [19], which have studied Web QoE exclusively for desktop browsers and desktop devices (or in some exceptional cases, emulating mobile devices), our measurement testbed consists of three different, non-emulated types of devices, including a smartphone device (Google Pixel 2 XL), a tablet (Google Pixel Slate), and a desktop computer (laptop), using WPT agents for Android and Linux. Chrome (last stable version) is used as browser.…”

Section: Web-qoe Datasets and Modeling Approachcontrasting

confidence: 64%

“…For example, the impact of multiple features such as transport protocols and network performance on PLT and AFT is studied in [15], based on a set of 244 million measurements collected during 6 months for the top-10000 Alexa websites. Other papers also measured the impact of similar features on PLT and SI or AFT in different countries and different types of networks [16], including mobile ones [17]. Most of prior work has stayed at the application level, which is problematic for ISPs, who have no direct access to in-browser metrics.…”

Section: Related Workmentioning

confidence: 99%

“…As a reference for performance evaluation, we begin by training and evaluating different ML models for the specific case of desktop measurements. Recall that desktop measurements represent the most common data source so far used in the Web-QoE literature [1], [2], [15]- [17], [19]. We then apply the trained models to both smartphone and tablet data, to highlight the cross-device lack of generalization and poor performance realized by single-device models in such multidevice scenarios.…”

Section: Desktop Models' Lack Of Generalizationmentioning

confidence: 99%

“…Conclusion: RUMSI inference and QoE prediction can be properly realized using CBD and session-based features, extracted directly from the stream of encrypted bytes. However, models so far proposed in the literature for single device types [1], [2], [15]- [17], [19] might not perform properly in the wild, where other devices than desktop machines are used for web browsing.…”

Section: Lack Of Generalization For Mobile Devicesmentioning

confidence: 99%

See 3 more Smart Citations

Are you on Mobile or Desktop? On the Impact of End-User Device on Web QoE Inference from Encrypted Traffic

Wassermann

Casas²,

Houidi³

et al. 2020

2020 16th International Conference on Network and Service Management (CNSM)

View full text Add to dashboard Cite

Web browsing is one of the key applications of the Internet, if not the most important one. We address the problem of Web Quality-of-Experience (QoE) monitoring from the ISP perspective, relying on in-network, passive measurements. As a proxy to Web QoE, we focus on the analysis of the wellknown SpeedIndex (SI) metric. Given the lack of applicationlevel-data visibility introduced by the wide adoption of end-toend encryption, we resort to machine-learning models to infer the SI and the QoE level of individual web-page loading sessions, using as input only packet-and flow-level data. In this paper, we study the impact of different end-user device types (e.g., smartphone, desktop, tablet) on the performance of such models. Empirical evaluations on a large, multi-device, heterogeneous corpus of Web-QoE measurements for the most popular websites demonstrate that the proposed solution can infer the SI as well as estimate QoE ranges with high accuracy, using either packetlevel or flow-level measurements. In addition, we show that the device type adds a strong bias to the feasibility of these Web-QoE models, putting into question the applicability of previously conceived approaches on single-device measurements. To improve the state of the art, we conceive cross-device generalizable models operating at both packet and flow levels, offering a feasible solution for Web-QoE monitoring in operational, multi-device networks. To the best of our knowledge, this is the first study tackling the analysis of Web QoE from encrypted network traffic in multi-device scenarios.

show abstract

Section: Web-qoe Datasets and Modeling Approachcontrasting

confidence: 64%

Section: Related Workmentioning

confidence: 99%

Section: Desktop Models' Lack Of Generalizationmentioning

confidence: 99%

Section: Lack Of Generalization For Mobile Devicesmentioning

confidence: 99%

See 2 more Smart Citations

Are you on Mobile or Desktop? On the Impact of End-User Device on Web QoE Inference from Encrypted Traffic

Wassermann

Casas²,

Houidi³

et al. 2020

2020 16th International Conference on Network and Service Management (CNSM)

View full text Add to dashboard Cite

show abstract

“…Even with object reordering, Furtak et al have shown that increased bandwidth improves page Speed Index by permitting large images to load more quickly [6]. However, others have observed that the impact of extra bandwidth is less pronounced on Page Load Time (PLT) [11]. Han et al have shown that extra bandwidth improves the performance of large pages [12].…”

Section: Related Workmentioning

confidence: 99%