SatNEx III - Future Web technologies and protocols over broadband GEO satellite networks

Gotta A., Fairthurst G., Secchi R., Vanelli Coralli A., Collina M., Luglio M., Roseti C., Sun Z., Cruickshank H., Caviglione L.

Satellite  SPDY  TCP 

This project explores the potential implications of introducing the HTTP/2.0 web technology on the design and operation of satellite networks. It has been recognised that as network bandwidth increases, the greater obstacle to the performance of web technology is latency rather than scarcity of capacity. HTTP/2.0 [RD-8] is a new protocol being design by the Internet Engineering Task Force (IETF) and World Wide Web Consortium (W3C) to address the limitations and lack of flexibility of the currently widely deployed HTTP/1.1 [RD-4]. The origin of the emerging HTTP/2.0 standard may be traced back to a GoogleTM suite of protocols, called SPDY. Proposed modifications to HTTP/1.1 include suppression of unnecessary Round Trip Time (RTT) delays to request and deliver web objects, multiplexing of web streams into a single connection, removal of the head of line (HoL) blocking, more compact encoding of HTTP headers, server initiated transmissions (Server Push), a new security model, and an extensible design. In parallel, a number of initiatives from well-known IT actors (GoogleTM, AppleTM etc) have proposed new transport methods, such as QUIC [RD-7] or Minion [RD-23], to enhance performance for web traffic. It is expected that the final HTTP/2.0 specification will replace HTTP/1.1 in every scenario HTTP/1.1 is currently used. The goal of Task3 of CoO3 Satnex-III was to understand the trajectory of the emerging web technologies and to evaluate the expected impact on satellite networking. Different aspects have been analysed by Satnex partners using both real satellite testbeds and emulation. This analysis included evaluation of SPDY performance over satellite and experiments to understand the expected interaction with PEPs (including scenarios with a SPDY proxy at a satellite gateway), the impact of security, and the effect of satellite capacity allocation mechanisms. The analysis also considered the impact of application protocols and the delay induced by end- system networks, such as a satellite-connected WiFi network. In general, experimental results show that HTTP/2.0 benefits the performance of web browsing via satellite by reducing latency and significantly lowering the overhead for several types of web pages. In particular, multiplexing a sequence of small web objects on a webpage onto a single connection enables siginicant gain and effectively reduces the HTTP overhead per object and per connection. The Server Push feature of HTTP/2.0, adds the ability for the server to send objects without explicit request, and was found particularly useful to reduce the download time. However these gains come at a cost, since they impact the design and operation of satellite networks. We also observed places where the tested implementations did not perform optimally, and were performance over satellite-connected WiFi networks was less than would be expected with HTTP/1.1. However, at the time of writing, the work on HTTP/2.0 is not complete and many mechanisms remain to be finalised. Modifications and adjustments are expected in future. Moreover, the current software support for HTTP/2.0 is not yet completely stable, and is expected to continue to evolve long after initial deployment. Satellite networking equipment will need to be updated to support HTTP/2.0. In particular, designs must re-consider the integration of PEPs withHTTP/2.0 and the updated security model, based on Transport Layer Security (TLS). TLS encryption prevents visibility of HTTP headers at satellite equipment, preventing use of current 23089/10/NL/CLP Executive and Summary Report 15/40 III CoO3 - Task 3 - Future Web Technologies and Protocols over Broadband GEO Satellite Networks HTTP acceleration PEPs. Our tests suggest that this problem may be mitigated by HTTP/2.0 itself, because some of the functions implemented in application-layer PEPs (compression, pipeling/multiplexing, push) are already present in HTTP/2.0. HTTP/2.0 is still in a development phase and it is important for the satellite community to watch carefully its progress and understand the implications this will place on the design of the higher layer packet processing. This is especially important because the new model does not introduce a single solution, but more represents a toolkit of techniques that will continue to evolve and will be deployed incrementally over a timescale of months and years ahead. There is also a need to understand the implications on network design/configuration, to ensure that satellite continues to offer high quality service comparable to terrestrial networks, and a need to re-consider the role of PEP in this new architecture. We also urge the satellite networking community to influence development of these standards. As network latency becomes important there is increased focus on low- delay networks, and it is important to ensure that design decisions (maybe unintentional) do not prejudice the performance over long delay paths when the new protocols are deployed.

Source: Project report, SatNEx III, 2014