WP3: Application: ICN for Green Video Sharing
Objectives of WP and expected results
The main objective of WP3 is to design a sharing framework for mobile video delivery in an ICN environment. Sharing can be realised in several different ways, which have been discussed in D3.1.1 of Task 3.1. By sharing network resources, but also the content itself, we attempt to reduce energy consumption of mobile devices. Given an ICN mode of operation, in-network caching resources are exploited in order to bring content closer to the end-users and reduce both network resource utilisation and delay. We have focused on a commuter train use-case, where users attempt to watch the same video content while commuting. This WP also focuses on the migration path from IP to ICN, taking also into account current trends in content delivery, i.e., Content Delivery Networks (CDNs).
The objectives of WP3, according to the DoW are as follows:
- Define a framework for collaboration and sharing incentives in order to achieve energy-efficient video delivery towards the access network.
- Investigate the ICN communication requirements for content delivery from content providers at the fixed and mobile environment.
- Define an architecture (based on pub/sub concepts) for communication in ICN environments.
- Specify the forwarding algorithm that enables efficient communication between fixed and mobile network devices, exploiting also inherent features of ICNs, such as in-network content caching.
- Exemplify migration paths from the current IPv4 Internet to ICN.
The expected result of WP3 (according to the DoW) is the definition of a “framework of ICN large scale pub/sub-based video sharing, IP migration path and in-network caching for mobile-oriented environments”.
The success criteria of WP3 (according to the DoW) are:
- Definition of algorithms for energy-efficient mobile video delivery taking also into account in-network caching resources.
- Definition of pub/sub forwarding and congestion control rules for content multiplexing in mobile ICN environments.
- At Least 20% Reduction of Power Consumption on normal days.
Below, we discuss the work accomplished within the first year of the project for each of the tasks of WP3.
Task 3.1 Framework
Summary of Y1
We have adopted a bottom-up approach for the design of the video-sharing framework and have identified the main components of the framework. We have built on three basic concepts of sharing, namely synchronisation of streams in caches for efficient video delivery on mobile devices, collaborative video streaming over 3G and P2P, and 3G offloading through Wi-Fi access points to provide initial context for the technical considerations of this design. Based on this we have identified the main building blocks of the framework, which we have discussed in detail in D3.1.1, namely video coding, descriptor-based Pub/Sub, flow control, multipath scheduling, cache management and optimisation, mobile-to-mobile communication management, routing and mobility management.
Summary of Y2
We continued investigation on framework components, in particular cache management and optimization, multipath forwarding strategies, and mobile-to-mobile communication management in the second year. To consolidate our video-sharing framework, relations among components are studied, e.g., video coding is considered in mobile-to-mobile communication protocol and the feature of video content is exploited in optimized caching.
In addition to the component study, power consumption of mobile networks, terminals, and applications are measured, and a theoretical framework for the assessment of the Quality of Experience of users streaming video in urban railway networks is proposed. They give foundations for energy saving and QoS guarantee in our video delivery environment.
Summary of Y3
We have completed the “Video Sharing Framework” by proposing a “QoE Assessment Framework”, which has been clearly missing from the related literature in the area. Our framework approximates the QoE that users get when streaming real-time audio and video in urban railway networks. In terms of video delivery, we have also investigated the option of aggregating Interest requests from end-user devices in order to reduce the energy consumption. These updated where reported in D3.1.3.
Task 3.2 Forwarding Strategy and IP Migration
Summary of Y1
In this Task we have tried to identify the migration options from an IP-based internet to a GreenICN one. The migration has to include the main content delivery players, which apart from the ISPs include CDN companies and their already operational infrastructure. The clean slate approaches introduced by most of the ICN-related projects to date have revealed a lot of the benefits of an ICN environment, but have also pointed to shortcomings of the present Internet architecture. Being clean-slate, however, they lack possible migration paths. In GreenICN, one of our first priorities is to provide a feasible migration path towards the shift to ICN Internet.
In Task 3.2 we have taken the first step on this direction. We have identified the challenges of a smooth migration to ICN and the integration of CDNs in an ICN environment. Our immediate next step is to incorporate our findings to the GreenICN Internames architecture.
Summary of Y2
We have improved COPSS to increase its efficiency as a publish/subscribe delivery system. It supports timely delivery, i.e, as soon as a publisher has new data, it can be sent to those who have expressed interest. We proposed a congestion control mechanism that allows COPSS to support large scale pub/sub, Video on Demand and Video streaming applications. The Congestion control proposed is able to fairly and efficiently utilize the bandwidth and avoid network congestion. We have also proposed a snippet/summary based delivery system that ensures that large video files are not pushed to the subscribers. Instead a snippet is sent and those interested can request for the full file (video on demand) or join a channel (video streaming). This reduces unnecessary traffic in the network, since large files that are not needed by the subscriber are not pushed.
We have also worked on the design and evaluation of a “Location-Independent Routing Layer” (LIRA), which is a proposal for a smooth migration from the IP Internet to an ICN-based future network. LIRA builds on top of IP and guarantees full backward compatibility, but also lowers the burden of deploying ICN functionality on top of the current network. In particular, we introduce the use of an extra layer in the OSI stack, a “3.5 layer” that sits on top of IP- and below the transport-layers. We introduce the concepts of “ephemeral names” and also “content provider-based name resolution”.
Summary of Y3
In terms of multimedia delivery to mobile users, we have proposed a WLAN multicast mechanism in order to improve the bandwidth efficiency of the wireless medium. An issue that has been overlooked in our previous studies is this of DoS prevention in ICN architectures. We have now proposed a solution for DoS prevention for the NDN architecture, which is reported in D3.2.2.
Task 3.3 Energy-Efficient Mobile Video Delivery and Caching
Summary of Y1
In this Task, we have focused on the difficulties faced by ICN in case of mobile environments. It is generally well-known that ICN brings benefits with regard to user mobility, that is, when an end-user moves to a different (part of the) network. Content and server mobility, however, still face the same issues as in an IP environment, where once the server (or content) moves to a different network, it is difficult for the network to resolve the hosted content. In this respect, we have carried out an extensive and in-depth evaluation of alternative ICN proposals and their reaction to mobility. We have identified the important (and positive) aspects that ICN achieves with regard to mobility and have also “flagged” the negative aspects. We will use this knowledge to build the mobility framework of the GreenICN Internames architecture.
In the context of Task 3.3, we have also focused on cache-aware routing and forwarding as per the DoW. We have designed and extensively evaluated a cache-aware routing scheme, based on the concept “hash-routing”. We have shown that off-path hash-routing brings great gains compared to on-path caching. We will use this knowledge and the related design to adjust our hash-routing algorithms to mobile environments too.
Summary of Y2
As a mechanism to support producer mobility as well as scalable routing, we proposed On-Path Resolver Architecture (OPRA). Hierarchically placed resolvers in OPRA support scalability of routing, minimize the overhead in handling binding update for mobile terminals, and provide opportunity to optimize route to migrated terminals.
Summary of Y3
In terms of the IP/CDN migration work, we have observed that some of the features of Information-Centric Networks can minimise load-imbalance in CDN server farms. This is an existing and pressing issue for CDN operators without any established solution. We have, therefore, investigated (both from a theoretical/analytical perspective and from a simulation one) techniques to reduce load imbalance by applying caching and chunking techniques. Our work is reported in D3.3.3. Finally, as part of Task 3.3, and in the context of the Video Sharing Framework of WP3, we have investigated cases where mobile video delivery has to be complemented by “helper nodes” to increase QoE and avoid playback disruptions. These “helper nodes” download content on behalf of the node that initiated the stream and is interested in the content.