Attila Mihály

Optimised local caching in cellular mobile networks

Motivated by the problem of increasing backhaul transmission costs in cellular mobile networks we examine the potential of highly distributed caching solutions. In particular we propose to overcome the problem of poor hit ratios in such solutions by forming caching domains which support pooling (the ability to fetch content from other caches when necessary) and equalisation (active transfers between caches to preload popular content). The characteristics of such a scheme are investigated by applying a new analytical model to a set of realistic examples. The model is verified against both computer based simulations and measurements in a real network, and it is found that the proposed scheme can reduce backhaul bandwidth requirements for an average cell inside a domain by a factor of 7 for web-like traffic and by a factor of 45 for video-like traffic, while the corresponding factors outside a domain amount to 41 and 997 respectively.

Prediction based energy efficient mobile positioning

Many smartphone applications or services require position information, e.g., location based services including mobile advertisement, mobile social network, etc. In this paper, we propose a prediction based mobile positioning mechanism which utilizes history mobility information in a smartphone. The proposed solution is implemented in an Android phone and has been evaluated in real-life for a period of three months. During the evaluation period the proposed mechanism has reduced energy consumption for positioning with 60% compared to GPS positioning. The accuracy is increased by around 76% compared to network based positioning only. Thus it provides significant benefit for those location-based services which needs longer time usage but less accuracy than GPS positioning.

The effect of incomplete sessions on peer-to-peer video on demand

Video on demand systems are going to be important services of the future Internet. To achieve scalability and fault tolerance, these systems should rely on distributed video delivery schemes, using peer-to-peer (P2P) networks built among the clients. However, generic P2P data sharing protocols cannot guarantee timely arrival of the video segments. Therefore, they must be adapted to video streaming by restricting P2P delivery into a download window. In this paper we investigate the effect of the frequently occurring phenomenon of incomplete sessions on the efficiency of stored video streaming via P2P data distribution. Through theoretical considerations and a simulation study, we’ve found that with a P2P download window the number of superfluous downloads can be limited, while retaining the efficiency of the P2P distribution.

Novel Availability Metrics for Network Topologies

This paper presents two novel definitions for availability concerning communication networks. It is shown that currently used network availability metrics cannot fully describe the resiliency of a network. The aim of this paper is to provide metrics that enable network operators to compare different network topology designs assuming a given population distribution. A transport network is considered to be in down state when at least a certain percentage of all subscribers for at least 10 seconds have no connectivity, or have only poor performance. We also present another availability metric named Session Completion Ratio, which is a more user-centric performance indicator since any user with poor performance or no connectivity deteriorates the overall network performance. Simulation results are presented and different transport network topologies are compared based on their availability performance, followed by a case study, where the availability of a network is improved by allocating extra capacity on the links. The results proved that single link failure-tolerant capacity setting leads to high-availability. Finally a tool is presented that helps identifying those parts of the network that have the strongest influence on resiliency.