Socket Intents: OS Support for Using Multiple Access Networks and its Benefits 
for Web Browsing

Tiesel, Philipp S.; Enghardt, Theresa; Palmer, Mirko; Feldmann, Anja

アイテム詳細

登録内容を編集ファイル形式で保存

一時保存へ追加

タグ情報を表示リリース履歴を表示詳細要約

公開

成果報告書

Socket Intents: OS Support for Using Multiple Access Networks and its Benefits for Web Browsing

MPS-Authors

/persons/resource/persons211491

Feldmann, Anja
Internet Architecture, MPI for Informatics, Max Planck Society;

External Resource

There are no locators available

Fulltext (restricted access)

There are currently no full texts shared for your IP range.

フルテキスト (公開)

arXiv:1804.08484.pdf
(プレプリント), 4MB

付随資料 (公開)

There is no public supplementary material available

引用

Tiesel, P. S., Enghardt, T., Palmer, M., & Feldmann, A. (2018). Socket Intents: OS Support for Using Multiple Access Networks and its Benefits for Web Browsing. Retrieved from http://arxiv.org/abs/1804.08484.

引用: https://hdl.handle.net/21.11116/0000-0002-158A-B

要旨

In today's Internet, mobile devices are connected to multiple access networks, e.g., WiFi/DSL and LTE. To take advantage of the networks' diverse paths characteristics (delay, bandwidth, and reliability) and aggregate bandwidth, we need smart strategies for choosing which interface(s) to use for what traffic. In this paper, we present an approach how to tackle this challenge as part of the Operating System (OS): With the concept of Socket Intents, applications can express what they know about their communication pattern and their preferences. Using our Socket Intents Prototype and our modified BSD Socket Interface, this information is used to choose the most appropriate path or path combination on a per message or per connection basis. We evaluate our system based on the use case of Web browsing: Using our prototype and a client-side proxy, we show the feasibility and benefits of our design. Using a flow-based simulator and a full factorial experimental design, we study a broad range of access network combinations (based on typical DSL and LTE scenarios) and real workloads (Alexa Top 100 and Top 1000 Web Sites). Our policies achieve performance benefits in more than 50% of the cases and speedups of more than factor two in 20% of the cases without adding overhead in the other cases.