British research about mobile P2P

Experimental Evaluation of the IP Multimedia Subsystem (IMS)

The IP Multimedia Subsystems (IMS) is considered by both network and service operators as a platform to bring about the long awaited all-IP convergence of the cellular world and the Internet.  IMS carries both signalling and bearer traffic in which multimedia sessions can be created, modified or deleted delivering voice, data and multimedia contents to end users (Figure 2).  As the number of devices that connects to the Internet via Third Generation (3G) networks increases, IMS has the advantage of providing Quality of Service (QoS), better billing system and integration of services which a 3G network per se cannot offer.

Given the substantial interest of IMS, we are carrying out a pilot study aimed at assessing its core functionality.  This is done by building a test-bed based on the specification of IMS as described by the Third Generation Partnership Project (3GPP) Release 5, which mandates the Internet Protocol version 6 (IPv6) for increased addressing space, Mobile IPv6 (MIPv6) for mobility, Internet Protocol Security (IPSec) for security and Session Initiation Protocol (SIP) for signalling and bearer traffic.  We integrate these as an overlay network allowing transparent connectivity between fixed and mobile networks over a multi-access network (LAN, WLAN, UMTS and GPRS) to emulate a typical IMS environment.  We carry out an experimental study to determine the level of maturity of the individual components, the interoperability of the components and how the platform would respond to issues like registration, mobility and handover in different environmental scenarios such  peer-to-peer file transfer, client-server services and voice services.  This project is strongly related to our study of pervasive, service-centric frameworks for advanced service provisioning (click here for further information).

Academic contact: Dr Antonio Liotta

 

Figure 2. The IMS architecture.

Mobile P2P Networking (PeerMob)

The Peer-to-Peer (P2P) computational paradigm is in essence an alternative to the well-know Client-Server (C-S) model.  While C-S applications and services involve a many-to-one relationship between application (client) and service repository (server), in the P2P model any host (or node) may simultaneously act as both client and server.  P2P computing represents a simple but extremely powerful paradigm shift which promises to facilitate the deployment of decentralised services (applications).

The full commercial deployment of P2P networks is hampered by a number of limitations of current systems.  In this project we address the following ones:

• P2P systems are intrinsically best-effort. It’s very hard, if not impossible, to determine the peer discovery time (i.e. the time needed for a peer acting in the role of a client to discover the relevant peer that will be acting as server). Also P2P systems are not geared to support any other form of Quality of Service.
• P2P systems are difficult to control. Since there is no central management authority in pure P2P systems, it is also very difficult to add management value to P2P services.
• P2P systems are difficult to secure and charge. These management functions are of paramount importance to the service provider.

The key goal of the project is to address these limitations in the particular context of mobile cellular networks. The key requirements of our mobile P2P framework are:

• To be able to run on thin mobile terminals (600Kbytes footprint including P2P core and application);
• To be macro handover resilient;
• To allow mobile P2P content distribution;
• To allows ‘deep’ mobile P2P semantic search/discovery;
• To feature disaster-recovery capability;
• To be is intrinsically scalable with number of users and amount of information published;
• To place the network operator in a unique position as provider of managed P2P services.

More specifically, we are looking at the integration of P2P networking protocols in the context of the IP Multimedia Subsystem (IMS) as depicted in Figure 3.  The project includes also the development of novel Mobile P2P services.

 Funding body: This project is funded by Vodafone R&D, U.K.

Academic contact: Dr Antonio Liotta

Figure 3. Mobile P2P networking over IMS.

Dynamic Network Clustering via Mobile Agents (completed)

Efficient clustering is a fundamental problem in the area of networking and distributed services.  Because of their peculiarities, autonomic systems are particularly sensitive to the way clusters are formed and cluster heads are elected.  Assuming that autonomic ubiquitous systems are composed of entities that move, attach/appear and disconnect/disappear fairly frequently, autonomic systems must rely on effective means for maintaining clusters and cluster heads.  This includes the ability to partition the system into an appropriate number of clusters (depending on the number of system entities) and elect the best possible cluster heads.

In this project, we have addressed the clustering problem for autonomic systems, presenting a novel approach that has the following advantages:

1)      It is based on a distributed algorithm that has a low (linear) cost (efficiency);

2)      It can satisfy precise constraints on the number of clusters (self-configuration);

3)      It creates provably near-optimal clusters and cluster heads (self-optimisation);

4)      It re-calculates near-optimal cluster heads in face of component failures or congestion (self-healing).

To the best of our knowledge, no existing technique satisfactorily addresses the combined requirements of efficiency, scalability, adaptability and optimality.  Our contribution includes an in-depth simulation-based analysis of the proposed approach, elaborating on its applicability to distributed monitoring, peer-to-peer systems, network overlays, application-level multicast, and content adaptation networks.  Figure 4 depicts our mobile agent based solution which uses agent cloning and agent migration to solve network clustering problem.

Academic contact: Dr Antonio Liotta

Figure 4. Network clustering algorithm based on a mobile agent system.

 

People

Academics: Dr Antonio Liotta

Research Assistants:

PhD Students: Marco Ballette; Ling Lin; Adetola Oredope.

Former group members: Carmelo Ragusa; Daniel H. Tyrode-Goilo.