Introduction to Network Crowdsourcing

New Challenges

The current wireless network service providers (NSP) face multifaceted challenges. On the one hand, the amount of traffics delivered over wireless networks continues to grow exponentially and will keep that way in the foreseeable future. One the other hand, the NSP’s profit made from traditional pipe-like services are not increasing in line with the traffic increase rate. Particularly, the super active over-the-top (OTT) service providers (e.g., Google, Amazon, Netflix, etc.) are revolutionizing the way people are using the network, and at the same time, undercutting the traditional business model of the NSPs. The most representative OTT service is the high-quality multimedia distribution, which is and will be the dominant wireless traffics. This trend is motivating NSPs to seek innovative service models and valuable partnership in response to the new economy drive by the OTT innovation.

Network Crowdsourcing

Figure 1. Process of Network Crowdsourcing

We propose a collaborative model between the OTT Content Provider (OCP) and wireless NSPs termed as Network Crowdsourcing, which is illustrated in Fig. 1. The essence of Network Crowdsourcing is to distribute content objects according to OCP’s requirements by soliciting distribution services from a group of NSPs, rather than from a unique NSP as in the traditional way. The NSPs are acting as contractors to undertake the task generated by the OCP, where each task should be the delivery of modularized content objects. Each NSP provides the OCP a service level agreement (SLA) showing how certain object could be delivered. The SLA should specify at least the following information: i) the network resources to be allocated to the object distribution, ii) the expense the OCP should spend for the object distribution, and, iii) the management functionalities can be provided to the OCP. Based on the SLAs offered by NSPs, the OCP may choose a group of NSPs to accomplish the content distribution task, therefore the notion of “network crowdsourcing”. The OCP can also provide task profiles specifying the preferences of each objects distribution subtask, including the amount of resource needed, budget and preferred management interfaces.

Enabling Technologies

The need of fully utilizing the network resources for content distribution has motivated the development of future network architectures based on the named data objects (NDOs), which is commonly known as the information-centric networking (ICN). As shown in Fig. 2, the paradigm of destination-based connection model should be transformed to content centric model. Modulation of contents and in-network caching are prevailingly applied to ICN architectures, which paves the way to highly efficient delivery of network service.

Figure 2. Change from connection centric to content centric

A fundamental feature that distinguishes the ICN model from the current Internet is that the information/content is decoupled from its sources, so that a particular content can be named and placed anywhere within the network. This feature not only facilitates the information distribution, but also provides the freedom in the creation of personalized, on-demand multimedia contents. For example, a popular video could be embedded with personalized advertisement tailored to the user’s preference, and patched with audio in a language suitable for users in an area. In this scenario, each of the components or the content objects, could be assigned with a name, delivered through a different path, and be combined towards the needs of different users.

Figure 3. Traditional Network vs. SDN

From the perspective of the NSP, the current trend towards software defined networking (SDN) technology provides the unique opportunity for the NSP to obtain more effective control and scheduling of its network resources. In particular, the SDN decouples the control plane and data plane of networking routers and switches, which enables the global control and management of the network fabric. The SDN controller can facilitate checking whether the required network resources are available; moreover, it can provide interfaces for applications to realize certain control and management functionalities.

Figure 4. Two-Tier SDN Controller Paradigm

Under the network crowdsourcing model, the OCP and clients could communicate with each other on which kind of network resources the client is able to utilize. For example, the client user may see multiple wireless networks owned by different NSPs, and each NSP may provide different kinds of access schemes such as broadcasting, WiFi and cellular. By aggregating such information, the OCP could understand the popularity of each object and the geographic distribution of clients served by each NSP, which could in turn greatly help the OCP generate the task profile. To accomplish the above mentioned functionalities, the design of two-tier SDN controller is introduced, as illustrated in Fig. 4. The negotiation protocol can work on the application layer interfacing the OCP and NSP’s network management application over the controller. After the negotiation process, the network distributes the modularized objects over the ICN. According to the popularity of each object and the location distribution of users, the objects will be pushed to and cached at corresponding parts of ICN, in order to facilitate wireless delivery through access networks.

Network Crowdsourcing based Content Delivery Network Architecture

Figure 5. Content Delivery Network

Fig. 5 exemplifies the application of Network Crowdsourcing to the content delivery system. The OCP needs to distribute 5 kinds of modularized objects in its server to users. With the negotiation and distribution process described above. Each object is cached at different locations with different amounts of copies, which complies users’ geographic distribution and their demands of each object. The objects can then be reassembled to form a playable content.

Advantages of Network Crowdsourcing

Due to the introduction of ICN paradigm, contents are cached in the network nodes based on their popularity, in addition, content delivery are more frequently carried out at network edge, which alleviates the traffic overhead and burden at core network and OCP server side. And content diversity is enabled owing to modularization of contents in ICN. Meanwhile, content delivery strategies are made in light of negotiation between OCP and NSP, such that the users’ historical statistics, including geographical distribution of users and their preferences, are made best use while NSPs can make full use of their available resource. Under such win-win circumstance, the allocation of limited network resources are optimized. What’s more, from users’ perspective, the requested contents are delivered through multiple channels, which enhances the throughput dramatically and efficiently use the access network bandwidth to improve QoS.