The reliable layered multicast using multiple sources and inter-source network decoding is proposed. The multi-source technique gives the path diversity that yields a better achiev-able data rate for layered transmissions. An optimization framework is formulated to find the optimal set of paths under several constraints. When the QoS guarantee of the
(a) (b) (c)
Figure 7.20: Visual example selected from video sequence “Bus” from various routing schemes: (a) Original, (b) QoS-aware w/ inter-source ND and multiple sources, and (c) QoS-oblivious w/o NC.
(a) (b) (c)
Figure 7.21: Visual example selected from video sequence “Football” from various routing schemes: (a) Original, (b) QoS-aware w/ inter-source ND and multiple sources, and (c) QoS-oblivious w/o NC.
(a) (b) (c)
Figure 7.22: Visual example selected from video sequence “Mobile” from various routing schemes: (a) Original, (b) QoS-aware w/ inter-source ND and multiple sources, and (c) QoS-oblivious w/o NC.
considering layered data cannot be achieved from the computed routes and some network resources are still available, a secondary source will find its optimal path and will transmit the same layered data to the considering destinations. This process will run repeatedly until the QoS guarantee is met or the network resources become unavailable. The network
coding is applied to the optimal set of paths. The inter-source network decoding is pro-posed to further improve the achievable data rate, whereas a sufficient number of network coded data are received not necessarily from the same source, the transmitted data can be recovered. Our experimental results show the improvement in the achievable data rate with QoS guarantee obtained from the proposed algorithm compared to previous works.
The simulations of scalable video coding show the gains in both subjective and objective qualities under various video sequences and network topologies.
Chapter 8
Conclusion
This chapter gives a discussion about our presented approaches and the con-clusion. Potential directions of future work are provided at the end of this chapter.
8.1 Discussion
The major part of link bandwidth in the Internet is occupied by content retrieval as a consequence of emerging content-oriented services and applications. In the process of content retrieval, a content delivery path often lies across multiple networks in the Internet, which can be categorized by location into the core and edge areas. In this dissertation, data transmission in both areas are facilitated by several novel content-oriented approaches to allow for efficient content delivery.
Data centers, which are major hosts of content servers, typically have direct connection to the core area of the Internet. The core area is therefore the first place where content starts a journey toward end-users. This area is prone to network congestion because it is loaded with a large amount of content retrieval traffic from end-users. Network caching has been an effective solution to network congestion. However, existing implementation of network caching, such as Web caches and Content Delivery Networks (CDNs), involves many unnecessarily complicated systems. Content-Centric Networking (CCN) architec-ture is considered to be a new solution to network congestion since it natively supports network caching in its protocol. Existing routing and caching schemes for CCN cannot
fully utilize in-network caches. We thus introduce a new cooperative routing scheme and a probabilistic caching scheme for CCN. The cooperative routing scheme optimizes data transmission paths based on the data similarity extracted from content retrieval statistics.
The probabilistic caching scheme lets CCN routers randomly cache data with a caching probability to reduce redundant copies of the same content cached by CCN routers along a transmission path. Both schemes cooperatively improve the utilization of in-network caches, leading to more efficient data transmission in the core area of the Internet. The content delivery completes the first part of data transmission in the core area and proceeds in the edge area.
In the edge area, multi-hop wireless networks would be prevalent as a result of the increasing number of wireless devices, accelerated by advancement in modern electronic and computing technologies. Unreliable quality and limited bandwidth are common issues in the transmission links of these networks. On the other hand, steaming bandwidth-demanding content over such networks has become the norm. Moreover, the quality of reproduced content at an end-user is susceptible to data loss. Quality-of-service (QoS) is thus necessary for data transmission in multi-hop wireless networks. We propose two new QoS-aware routing schemes for multi-hop wireless networks, one is for unicast and the other is for multicast. In essence, our proposed routing schemes select optimal routes in given networks according to QoS requirements and the importance of data. Network coding techniques are exploited to further improve network resource utilization. Our routing schemes not only better guarantee QoS of data transmission in multi-hop wireless networks than several existing schemes, but also achieve efficient utilization of network resources.
We propose several content-oriented approaches for efficient data transmission in the Internet. Results from extensive evaluation show that our schemes can address data trans-mission more efficiently than several existing schemes which ignore the properties of data.
Our proposed schemes may be orthogonal, considering that they are applicable to networks in different areas. Nevertheless, they can be jointly used to enable efficient content deliv-ery over heterogeneous networks in the future Internet. For instance, a new framework for efficient content delivery can be described as follows. We can utilize CCN architecture along with the cooperative routing and probabilistic caching schemes to facilitate data
transmission in the core area. Much of redundant traffic could be eliminated by efficient in-network caches, thus network congestion would rarely happen. Upon receiving the data from the core area, networks in the edge area forward them to end-users. Some of data may be lost in the this area before reaching end-users as a result of unreliable wireless links and limited link bandwidth. In this case, our QoS-aware routing schemes can be used both to achieve a QoS guarantee and to prioritize data transmission. The cooperation among all of our proposed schemes would enable efficient content delivery in the future Internet.
It is worth to note that our work has some limitations. For example, our Optimal Cooperative Routing Protocol (OCRP) is driven by exchanges of signalling packets be-tween CCN routers. Excessive signalling packets could affect the performance of particular CCN routers and routing stability. The number of signalling packets created depends on several parameters, such as the number of CCN routers, threshold of popularity index, and traffic sampling interval. These parameters are unique to each network and are not necessarily the same as those used by this work. In the performance analysis of a prob-abilistic caching scheme, we model the requests of content with Independent Reference Model (IRM). However, recent studies in [118] point out that IRM may reduce the accu-racy of caching analysis because temporal locality of requests is ignored. We also assume Zero Download Delay (ZDD) in this work. In fact, download delay occurs for every cache-miss and may contribute to the discrepancy between the analytical results and the actual performance in real networks. Next, let us consider our QoS-aware routing schemes for multi-hop wireless networks. In our routing scheme for unicast transmission, the optimal route selection and CNC establishment are separate steps. An optimal route for the entire unicast sessions is calculated in the first step. Upon the obtained route, our Cooperative Network Coding Establishment (CNCE) algorithm decides whether CNC will be applied at which intermediate node. If this optimal route does not match any CNC structure, the network will not set up any CNC. To improve the opportunity of CNC, the optimal route calculation should be aware of CNC structures and QoS requirements. However, this would cost much more complexity than our proposed scheme. Finally, in our work on QoS-aware routing scheme for multicast transmission, we assume that only a multicast session exists in a multi-hop wireless network. In practice, more than one session may oc-cur simultaneously and may increase contention for network resources. Hence, a multiple
source technique would be limitedly used in some networks.