Peer-to-peer (P2P) applications are enormously popular on the Internet. Their uses vary from file sharing to Voice–over–IP to gaming and more. Increasingly, users are moving toward wireless networked devices and wish to continue using P2P applications in these new environments. This project is interested in investigating the use of P2P communications in order to enhance service delivery, both to the wired and the wireless end user. This use promises more robust service delivery platforms that are also much less demanding when it comes to network architecture and resources. The project proceeds in two directions. The first targets a generalized implementation of P2P technologies in wireless systems, enabling them to operate over the very particular architecture of Mobile Ad hoc Networks (MANETs). The second takes on the challenge of streaming video on the Internet, and studies the application of P2P to enhance user experience.

A Framework for Peer-to-Peer Computing in Mobile Ad hoc Networks

In our work, we developed a framework for practical mobile P2P computing in MANETs (P2P–MANETs). Though P2P and MANETs share some similarities, such as self-organization, dynamism, and resilience to failure, it is necessary to create new P2P algorithms that take advantage of the realities of MANETs. These algorithms must account for the numerous challenges found in these networks, including node mobility, resource constrained nodes, and the necessity of fully distributed algorithms. 

Our work includes several components. First, mobile nodes must be able to locate and join the P2P overlay. Next, the overlay peers must form a topology of connections between themselves. As it is important that peers contribute to the network by sharing their resources and forwarding traffic for others, we there developed for the third component a dynamically priced incentive scheme which rewards users for contributing to the network. We also developed a path selection algorithm to allow peers to select how many parts of a file to download from specific servers and which paths to choose in order to satisfy the user’s preference for download time and cost. For the final component of our framework, we developed a content distribution system that allows users to download large files through the use of network coding and multicasting. We examined the performance of our design in simulation and determined that it clearly outperformed existing approaches. Further, we implemented our design in Java and verified its functionality in a small network, confirming its real world utility.

Peer-to-Peer Video Streaming on the Internet

Large-scale video streaming on the Internet is challenging due to its high bandwidth and stringent delay requirements. Meanwhile, P2P networks have drawn particular attention from both the research community and the industry due to their low cost and high scalability. P2P video live streaming schemes can be classified into push schemes that construct multicast trees and pull schemes that use the swarming technique. The former is vulnerable to peer churn while the latter has a long playback delay. We devised a network-driven push-pull hybrid scheme for large scale video live streaming applications that utilizes the network resources efficiently, has a shorter delay than existing hybrid schemes, and is potentially more robust than all the existing schemes. 

We implemented a discrete-event simulator to evaluate the scheme. The results show that the scheme, depending on the settings of the substrate Internet, generally has a small path stretch (less than 2), which is similar to the small and medium scale push schemes and several timers lower than existing pull and push-pull schemes. Even if we use our neighborhood overlay construction scheme for existing hybrid schemes, the comparative push-pull hybrid scheme use 69% more network resources because it cannot build a short tree as our routing protocol.

This project looks at enhancing the user experience through utilizing elements of user spaces and ambiences and involving them directly into network operation. The result service delivery architectures would be intuitively adaptive not just to user mobility and application requirements, but also to user’s personal interests and background. For example, it is possible to link user profiles in their social networks with their communication and mobility patterns. In this manner, user environments (ads, services, offers) would react to user availability and availability of people they know within the same locale. Such enhancements would enrich not just the user’s networking experience, but extends to the user’s overall experience. At the same time, the increasing use of the Web for everyday tasks is making mobile Web services an essential part of the Internet customer's daily life. To get the most relevant Web services that fulfill the user's request, the user has to construct the request using the keywords that best describe the user's objective and match correctly with the Web Service name or location. Clustering Web services based on function similarities would greatly boost the retrieval capability of Web services search engines. 

A Framework for Smart Social Spaces

Stimulated by the recent popularity of wireless mobile communications as well as the increased availability of location-aware services and contextual information, pervasive computing aims at providing “what you want, when you want it, how you want it, and where you want it” services to users and applications. Although sensor networks can provide useful contextual information, they usually provide information about the physical environment but are incapable of accurately predicting social information such as the interests and preferences of users within the environment. There is a wealth of information within social networks, which we use to provide a whole new level of mobile social environments and targeted contextual services. We use this information to create a smart adaptive wireless environment. This environment retrieves social information about users within its coverage from Social Networks. It then uses the aggregate of this information to provide services within the domain of the environment that cater to the preferences and interests of these users. These services are usually provided within the context of a smart space, which can be defined as an environment that is capable of acquiring and applying knowledge about the environment and its inhabitants in order to improve their experience in that environment. Our system is capable of identifying users and their personal profiles and then imports their social context into a system through standardized formats.

By introducing IMS into a smart space architecture a new session control layer is formed. The session control layer includes Call Session Control Functions (CSCFs), which provide the registration of the endpoints and routing of the SIP signaling messages to the appropriate application server. The CSCF communicates with the transport layer to guarantee Quality of Service across all services. This session layer also includes the Home Subscriber Server (HSS) database that maintains the unique service profile for each end user. The end user’s service profile stores all of the user service information and preferences in a central location. Unifying this information, applications can share and exchange information to form centralized personal directories, enable features like presence and context awareness, and blend together services. This structured framework also makes the administration of user data significantly simpler and insures network consistency amongst all service subscribers. 

Clustering for Mobile Web Services Discovery

Effective Web service discovery is an important issue, especially for mobile Web services, where users are always on the move and finding Web services is crucial to their highly dynamic needs.  Mobile Web services, in one of their architectures, use the traditional Universal Description and Discovery and Integration (UDDI)-based Web service discovery. UDDI matching approach lacks the ability to recognize the content of the Web service description file. We developed an approach to improve service discovery of Web services by clustering similar services through mining WSDL documents. We identified five key features that are extracted and integrated in order to group Web services into functionality-based clusters. 

Our clustering approach can be integrated into search engines to improve the quality of Web service discovery by helping to identify the Web services relevant to a user request. This will, in turn, add value to the discovery process by providing users with better quality options in selecting a best-fit service. Experiments show a performance improvement in the quality of the retrieval compared with previous approaches. We also incorporate users’ context in the matching process and rank the retrieved Web services for mobile users based on best-fit to their context to facilitate the selection decision.

The emerging Broadband Wireless Networks (WBN), such as 3GPP’s Long Term Evolution (LTE) and IEEE’s 802.16 (WiMAX), come to answer a multifaceted and serious need to extend broadband Internet delivery to the wireless end users. With increasing demand from various sectors, ranging from civilian to commercial to governmental and services, the push for WBN infrastructures has never been greater. Our interest in this project is to facilitate the design of profitable and efficient BWN networks while reducing both economic and operational considerations. 

Our research covered four complementary directions. The first direction investigates the design of new radio resource management frameworks that are more fit to the emerging characteristics of BWNs. The second direction establishes a computation framework to analyze interference and capacity in BWN, especially those employ dedicated wireless relay stations to either enhance network capacity or extend coverage. Such features are achieved economically as opposed to increase the number of base stations and their wired connectivity. The third direction looks at the operation of multihop relay station in BWN, and how the design of the frame structure (which dictates the allocation of radio resources to the various users and relay stations) affects network performance. Scheduling, which deals directly with user allocations in a manner that satisfies both network and user requirements, makes the scope of the fourth direction where we investigate the general tradeoffs between operational complexity and objective satisfaction. The motivation for the fourth direction is the design of operationally efficient schedulers for broadband networks. 

Interference and Capacity Analysis in Mulithop Relay Networks

The high level objective of this work is to facilitate sound analytical basis for designing BWN employing relay stations. A direct challenge in realizing this objective is the nature of OFDM technology in that, despite increasing network robustness against certain medium challenges, requires careful considerations at the design stages to circumvent certain effects. Such effects include high-power-amplifier nonlinearity at the transmitter side and Doppler Effect during the transmission process over the channel. In a multihop setting, these effects cumulate to an extent that unjustifiably devastates network performance in terms Bit Error Rate Degradation. Our analysis thus far confirmed this, with the analysis being confirmed by a system level simulation.

Further work expanded on these findings to work exploring the possibility of reducing the Doppler effect on the multihop relaying channels using directional antenna at both the transmitter and the receiver sides, and devising analytical models to capture the effect of subcarrier collision when the occupied subcarriers are simultaneously used at different relay stations. We are currently working on harmonizing the resulting computational framework for the capacity and interference, which will provide network designers and operators with great flexibility as it will be able to accommodate various deployment scenario

Frame structures for Multihop Relay Networks

There are two main types of multi-hop relay stations: transparent, which are not able to transmit control information, and non-transparent, which have the capability to transmit such information. Our work focuses, mainly, on non-transparent relay stations due to their complexity and ability to operate in a more than two hop environment. Currently, the latest IEEE amendment provides two different frame structures – single and multi frame – to allocate resources in a multi-hop relay environment. We examine the effect of the frame structure type on WiMAX network employing non transparent relay stations (ntRSs). Specifically, we examined the effect of the frame structure type on various operational metrics. To facilitate this examination, we developed an add-on for NS-2 by modifying the light WiMax (LWX) add-on contributed by Lai and Chen. We expanded the LWX to consider ntRS as the LWX was only aimed at evaluating setups with transparent RS (tRS). Our modifications include the implementation of the frame structure as described by the IEEE 802.16j, extending support for more than two hop relay connections, as well as making changes to the uplink and downlink scheduler.

To account for real life network deployment scenarios, and at the same time trying to be as generic as possible; we have chosen the star and line network topologies, to evaluate the two proposed frame structures. Our findings indicate that there are specific advantages both single frame and multi-frame have to offer. The single frame structure has shown that is able to achieve a better performance on average compared to the multi-frame one. The multi-frame structure; however, is able to support a higher number of users.

Uplink Scheduling in BWNs

The intent of this project was to investigate the design of uplink scheduling in BWNs. Specifically, our focus has been on the uplink scheduling of the emerging BWN from 3GPP, namely Long Term Evolution (LTE), which uses SC-FDMA as a basic for multiple access in the uplink. SC-FDMA differs from OFDMA in facilitating better channel performance and long UE battery life. However, it shares a similarity of exploiting the simultaneous transmission of multiple carriers to overcome certain medium challenges. The first part of the project involved an investigation of uplink scheduling proposals that have been made in the general literature and industry. Such schedulers employing an LTE feature of adapting transmission rate and quality to channel conditions and are hence often called Channel Dependent Schedulers (CDS). The basic frame and transmission setup of an SC-FDMA frame results in a complex scheduling problem to realize an optimal scheduling. This is directly a result of handling two domains (time and frequency) at the same time. Through our investigations, however, we were able to realize that operationally efficient schedulers can be realized without adversely affect service delivery. Our investigation of schedulers include QoS and ARQ/HARQ consideration, in addition to the support of features such power control and MIMO transmissions. 

Novel Frameworks for Resource Management in BWNs

A strong feature of BWNs, enabled by the thrust towards all-IP network cores, is that they take advantages of different connectivity options in terms of access technology (LTE, WiMAX, IEEE 802.11, etc.), network structure (single-to-many, relay, femtocells, etc), and spectrum access. The resultant network is a composite that merges – where and when applicable – the advantages of the individual option. To realize the full strength of this composite, however, requires novel radio resource management frameworks as traditional frameworks were not designed to accommodate such heterogeneity Based on our preliminary investigation, we discerned three main principles in framework design: (1) reducing the operational cost of resource management modules; (2) jointly manage the resources of the different connectivity options, both in the short term (scheduling, admission control) and in the long term (reservation); (3) devise opportunistic modules that take advantage of new network features. Towards this latter point, we introduced the notion of Operator Driven Handovers (ODHs) whereby an operator utilizes handovers between connectivity options to enhance network operation and achieve certain objectives. Specifically, we employed ODHs to reduce the operator’s monetary cost of service delivery, periodically adjusting user associations while strictly upholding their service level agreements. We also explored the definition of resource reservation in networks where spectrum can be traded between operators or through a spectrum exchange, and outlined how the relevant economic considerations are to be infused in the design of the resource management framework.