We investigate the potential of different pre-fetching and/or caching strategies for different user behaviour with respect to surfing or browsing in a catch-up-TV network. To this end we identify accounts and channels associated with strong or weak surfing or browsing respectively and study the distributions of hold times for the different types of behaviour. Finally we present results from a request prediction model and a caching simulation for the different types of behaviour and find that the results are relatively similar.

All forecasts of Internet trac point at a substantial growth over the next few years. From a network operator perspective, efficient in-network caching of data is and will be a key component in trying to cope with and profit from this increasing demand. One problem, however, is to evaluate the performance of different caching policies as the number of available data items as well as the distribution networks grows very large.

In this work, we develop an analytical model of an aggregation access network receiving a continuous flow of requests from external clients. We provide exact analytical solutions for cache hit rates, data availability and more. This enables us to provide guidelines and rules of thumb for operators and Information-Centric Network designers.

Finally, we apply our analytical results to a real VoD trace from a network operator and show that substantial bandwidth savings can be expected when using in-network caching in a realistic setting.

Video content, of which YouTube is a major part, constitutes a large share of residential Internet traffic. In this paper, we analyse the user demand patterns for YouTube in two metropolitan access networks with more than 1 million requestsover three consecutive weeks in the first network and more than 600,000 requests over four consecutive weeks in the second network.

In particular we examine the existence of “local interest communities”, i.e. the extent to which users living closer to each other tend to request the same content to a higher degree, and it is found that this applies to (i) the two networks themselves; (ii) regions within these networks (iii) households with regions and (iv) terminals within households. We also find that different types of access devices (PCs and handhelds) tend to form similar interest communities.

It is also found that repeats are (i) “self-generating” in the sense that the more times a clip has been played, the higher the probability of playing it again, (ii) “long-lasting” in the sense that repeats can occur even after several days and (iii) “semi regular”in the sense that replays have a noticeable tendency tooccur with relatively constant intervals.

The implications of these findings are that the benefits from large groups of users in terms of caching gain may be exaggerated, since users are different depending on where they live and what equipment they use, and that high gains can be achieved in relatively small groups or even for individual users thanks totheir relatively predictable behaviour.

Streaming video continues to experience unprecedented growth. This underscores the need to identify user-centric performance measures and models that will allow operators to satisfy requirements for cost-effective network dimensioning delivered with an acceptable level of user experience. This paper presents an analysis of two novel metrics in the context of fixed capacity links: (i) the average proportion of a video’s playing time during which the quality is reduced and (ii) the average proportion of videos which experience reduced quality at least once during their playing time, based on an M/M/∞ system. Our analysis is shown to hold for the more general M/G/∞ system for metric (i), but not for (ii) and simulation studies show an unexpected form of sensitivity of metric (ii) to the flow duration distribution, contrary to the norm of increasing variance causing worse performance. At typical operational loads these new metrics provide a more sensitive and information rich guide for understanding how user experience degrades, than the widely used average throughput metric does. We further show that only the combined use of this existing and our new metrics can provide a holistic perspective on overall user performance.

The coverage areas of cellular networks are logically subdivided into service areas. Each service area has a local anchor node which “hides” the mobility inside the area and the entire network has a global anchor node which “hides” the mobility between areas.

The concept of unique local anchor nodes per service area was invented to simplify routing but has been found to complicate expansion. The rapidly growing demand for cellular access has therefore prompted for alternative solutions with pools of local anchor nodes per service area. Such pools are now deployed by several operators all over the world.

Users in pooled service areas are mapped to specific pool members according to a load distribution policy, but the mapping can change as a result of node failures or operator interventions. Such changes take a certain time to implement and cause additional load on the anchor nodes. We study these processes in detail and derive closed form expressions which allow operators to control the trade-off between rapid changes and acceptable loads.

Finally we show that the key assumptions of our model are in agreement with measured data and demonstrate how the model can be applied to investigate the effects of different network settings (timers) under different user behaviour (traffic and mobility).

Contrary to current solutions to this problem, which typically are slow and/or inaccurate, our results enable fast and accurate analysis of different scenarios thereby enabling operators to maximise utilisation of the existing investments and at the same time avoid potentially dangerous situations of overload.

Quantifying quality of experience for network applications is challenging as it is a subjective metric with multiple dimensions such as user expectation, satisfaction, and overall experience. Today, despite various techniques to support differentiated Quality of Service (QoS), the operators still lack of automated methods to translate QoS to QoE, especially for general web applications.

In this work, we take the approach of identifying unsatisfactory performance by searching for user initiated early terminations of web transactions from passive monitoring. However, user early abortions can be caused by other factors such as loss of interests. Therefore, naively using them to represent user dissatisfaction will result in large false positives. In this paper, we propose a systematic method for inferring user dissatisfaction from the set of early abortion behaviors observed from identifying the traffic traces. We conduct a comprehensive analysis on the user acceptance of throughput and response time, and compare them with the traditional MOS metric. Then we present the characteristics of early cancelation from dimensions like the types of URLs and objects. We evaluate our approach on four data sets collected in both wireline network and a wireless cellular network.

In an increasingly competitive environment it is more important than ever for operators to keep their end users satisfied. User satisfaction is often characterised in terms of Quality of Experience (QoE), a subjective metric with multiple dimensions such as expectations, content, terminal, environment, cost and performance. QoE is typically quantified as MOS, mean opinion score, which is obtained by averaging the ranks of a number of voluntary users for controlled combinations content/terminals/performance etc. While this approach has many advantages, there are also a number of difficulties such as representativeness (the number of users as well as the number of objects and devices all have to be kept small); validity (the results may be biased by the situation, the setting, the renumeration and so on); and applicability (it is not clear how different numbers map to notions such as “acceptable” or “unacceptable” and operators alone cannot do very much about factors such as content).

We thus investigate the possibilities of detecting user opinions in the above, simplified, terms and from the network itself; with actual expectations, content, terminals, environments, costs and performance for virtually all users all the time. To this end we revisit the earlier suggestion that user opinions be reflected in their behaviour such that poor performance may result in interrupted requests. These works have, however, considered single flows hence we extend that idea to web pages which are groups of flows. In this paper we present our methods to group flows, interpret users, and characterise performance and we make a first assessment of the correlations between web page interruptions and network performance characteristics.

Mobile internet has been widely adopted and it is expected to rise to almost 4 billion users by 2020. Despite the research effort dedicated to the enhancement of its performance, there still exists a gap in the understanding of how TCP and its many variants work over LTE. To this end, this paper evaluates the extent to which five common TCP variants, CUBIC, NewReno,Westwood+, Illinois, and CAIA Delay Gradient (CDG), are able to utilise available radio resources under hard conditions, such as during start-up and in mobile scenarios at different speeds. The paper suggests that CUBIC, due to its Hybrid Slow-Start mechanism, enters congestion avoidance prematurely, and thus experiences a prolonged start-up phase. As a result, it is unable to efficiently utilise radio resources during shorter transmissionsessions. Besides, CUBIC, Illinois and NewReno, i.e., the loss-based TCP implementations, offer better throughput, and are able to better utilise available resources during mobility than Westwood+ and CDG - the delay-based variants do.

Internet traffic from a fibre based residential access network is investigated concerning traffic volumes and link load. Also the cost of the services is analyzed. We show that 1 Mbps accesses subscribers maintain high loads, and that the price they pay per GB used is five times higher than the one paid by 100 Mbps access subscribers.

TV-on-Demand services have become one of the most popular Internet applications that continuously attracts high user interest. With rapidly increasing user demands, the existing network conditions may not be able to ensure a low start-up delay of video playback. Prefetching has been broadly investigated to cope with the start-up latency problem, which is also known as user perceived latency. In this paper, two datasets from different IPTV providers are used to analyse the TV program request patterns. According to the results, we propose a prefetching scheme at the user end to preload videos before user requests. For both datasets, our prefetching scheme significantly improves the cache hit ratio compared to passive caching and we note that there is a potential to further improve prefetching performance by customizing prefetching schemes for different video categories. We further present a cost model to determine the optimal number of videos to prefetch. We also discuss if there is enough time for prefetching. Finally, more factors, which may have an impact onoptimizing prefetching performance, are further discussed, such as the jump patterns over different time in a day and the the distribution of each video’s viewing length.

TV-on-Demand service has become one of the most popular Internet applications that continuously attracts higher user interests. With rapidly increasing user demand, the existing network conditions may not be able to ensure low start-up delay of video playback. Prefetching has been broadly investigated to cope with the start-up latency problem which is also known as user perceived latency. In this paper, we analyse request patterns for TV programs from a popular Swedish TV service provider over 11 weeks. According to the analysis, we propose a prefetching scheme at the user end to preload videos before user requests. Our prefetching scheme significantly improves the cache hit ratio compared to terminal caching and we note that there is a potential to further improve prefetching performance by customizing prefetching schemes for different video categories. We further present a cost model to determine the optimal number of videos to prefetch. Finally, we discuss available time for prefetching and suggest that when to make prefetching decisions depends on the user demand patterns of different video categories.

Data volumes in communication networks increase rapidly. Further, usage of social network applications is very wide spread among users, and among these applications, Facebook is the most popular. In this paper, we analyse user demands patterns and content popularity of Facebook generated traffic. The data comes from residential users in two metropolitan access networks in Sweden, and we analyse more than 17 million images downloaded by almost 16,000 Facebook users. We show that the distributions of image popularity and user activity may be described by Zipf distributions which is favourable for many types of caching. We also show that Facebook activity is more evenly spread over the day, compared to more defined peak hours of general Internet usage. Looking at content life time, we show that profile pictures have a relatively constant popularity while for other images there is an initial, short peak of demand, followed by a longer period of significantly lower and quite stable demand. These findings are useful for designing network andQoE optimisation solutions, such as predictive pre-fetching, proxy caching and delay tolerant networking.

Today, cellular networks rely on fixed collections of cells (tracking areas) for user equipment localisation. Locating users within these areas involves broadcast search (paging), which consumes radio bandwidth but reduces the user equipment signalling required for mobility management. Tracking areas are today manually configured, hard to adapt to local mobility and influence the load on several key resources in the network. We propose a decentralised and self-adaptive approach to mobility management based on a probabilistic model of local mobility. By estimating the parameters of this model from observations of user mobility collected online, we obtain a dynamic model from which we construct local neighbourhoods of cells where we are most likely to locate user equipment. We propose to replace the static tracking areas of current systems with neighbourhoods local to each cell. The model is also used to derive a multi-phase paging scheme, where the division of neighbourhood cells into consecutive phases balances response times and paging cost. The complete mechanism requires no manual tracking area configuration and performs localisation efficiently in terms of signalling and response times. Detailed simulations show that significant potential gains in localisation efficiency are possible while eliminating manual configuration of mobility management parameters. Variants of the proposal can be implemented within current (LTE) standards.

Today, cellular networks rely on fixed collections of cells (tracking areas) for handset localisation. This management parameter is manually configured and maintained and is not regularly adapted to changes in use patterns. We present a decentralised approach to localisation, based on a self-adaptive probabilistic mobility model. Estimates of model parameters are built from observations of mobility patterns collected online using a distributed algorithm. Based on these estimates, dynamic local neighbourhoods of cells are formed and maintained by the mobility management entities of the network. These neighbourhoods replace the static tracking areas used in current implementations by using the tracking area list facility of LTE. The model is also used to derive a multi phase paging scheme, where the division of cells into consecutive phases is optimal with respect to a set balance between response times and paging cost. The approach requires no manual tracking area configuration, and performs localisation efficiently in terms of number of location updates, page

In this paper we study traffic patterns in a large municipal WiFi network and in particular those of the most bandwidth hungry application, viz. YouTube, for which we provide a detailed analysis of demand in different geographical areas and over time. We consider the possibilities to reduce network traffic and increase Quality of Experience (QoE) by serving repeated requests for YouTube videos from caches placed either at the network head end, at the wireless access points, or in the user devices. Our data confirms that a significant part of the YouTube traffic can be served by such devices and that there exists a potential to optimize caching performance by exploiting the content demand locality. We also discover a previously unknown pattern of periodicity in content demand and present a simple example of how to exploit this in cache design.

In this work, we study YouTube traffic characteristics in a medium-sized Swedish residential municipal network that has ~ 2600 mainly FTTH broadband-connected households. YouTube traffic analyses were carried out in the perspective of video clip category and duration, in order to understand their impact on the potential local network caching gains. To the best of our knowledge, this is the first time systematic analysis of YouTube traffic content in the perspective of video clip category and duration in a residential broadband network. Our results show that the requested YouTube videoclips from the end users in the studied network were imbalanced in regarding the video categories and durations. The dominating video category was Music, both in terms of the total traffic share as well as the contribution to the overall potential local network caching gain. In addition, most of the requested video clips were between 2-5 min in duration, despite video clips with durations over 15 min were also popular among certain video categories, e.g. film videos.

In this paper we describe a systematic study on long-term evolution of residential broadband Internet traffic covering 5 calendar years from June 2007 to May 2011. The traffic evolution is characterized both in the term of the total traffic volume, as well as the traffic volumes and shares for different application categories (file sharing, video streaming etc.), with the focus on comparing the traffic on the per IP user basis and among different broadband subscription groups. The results show that the average daily total traffic generated by each private end user increased only by about 33 % during the past 5 years. Further, the results show that the P2P filesharing has been dominating the network total traffic, but the daily file-sharing traffic volume per end user largely remains the same. Also, the daily streaming-media traffic volume per end user has increased dramatically by over 500% during the studied period of time. In the meantime, the daily web-browsing traffic volume per end user has increased by about 300%. Finally, a further investigation among 4 different FTTH broadband subscription groups with 1, 10 , 30, and 100 Mbit/s symmetric access speeds shows that the lower the access speed, the more diversified the end user traffic tend to be.

In this work, the performance of 5 representative caching replacement policies was investigated and compared for caching Internet video-on-demand (VoD) in local access networks. Two measured traces of end-user requests were used in the analyses for two typical VoD services: TV-on-demand and user generated content represented by YouTube. The studied policies range from simple least recently used (LRU) and least frequently used (LFU) algorithms to more advanced ones denoted as LFU-dynamic lifespan (LFU-DL), Adaptive replacement cache (ARC) and Greedy-dual size frequency (GDSF). Our results show that the ARC policy always outperforms the other policies due to its adaptive nature and its ability to track changes in the traffic patterns. On the other hand, the simple LRU policy can also achieve a caching performance which is comparable to that of the more advanced ARC policy especially for the TV-on-demand service when the potential caching gain is high. On the contrary, the simple LFU policy always shows the poorest performance.However, by applying a proper lifespan supplement under the LFU-DL policy, the caching performance can be effectively enhanced to the level achievable using ARC and LRU policies. Moreover, the GDSF policy does not outperform simple LRU or LFU-DL, especially for YouTube video clips when the potential caching gain is relatively low. The advantage of GDSF manifested in our analysis is, however, its outstanding cache space usage efficiency among the five studied caching algorithms.

Distribution of media data over the Internet is increasing in popularity and volume. This poses challenges not only for network operators but also for service providers when it comes to serving the demand in a cost-efficient way. In this paper, we approach this problem by investigating the potential of co-operative approaches where locality in space (users in the same network) and locality in time (concurrent downloads) are exploited such that as many requests as possible may behandled inside the access and metro networks. This approach may contribute not only to reducing transport costs (less traffic in core networks and at peering points) by but also improve the end user experience (by reduced round trip times and exclusion of some possible bottlenecks). To this end we develop a method to measure the possible gains from, firstly, optimal handling of concurrent downloads and, secondly, optimal utilization localavailability. We apply the method to BitTorrent data from two metropolitan access networks and find that the bandwidth savings amount to between 10% and 20% when optimizing concurrent downloads and between 56% and 66% when exploiting local availability with a simulated network cache.

This paper presents our work regarding the comparison of the behaviour of some of the major TCP variants in a LTE network. Using the LENA LTE simulator and the DCE framework, the behaviour of the Linux implementations of seven TCP variants are compared. The evolution of the throughput, congestion window and queuing delay are studied for four scenarios with different network loads and flow types. Our measurements show that, in the situations we consider, most variants are able to quickly reach full link utilisation. However, to reach the same throughput, they create different amounts of queuing delay. On the one hand, loss based algorithms tend to totally fill the queue, creating huge queuing delays and inducing packet losses. On the other hand, delay based variants manage to limit the queue size and decrease the amount of packet dropped by the eNodeB but struggle to reach the maximum throughput in some circumstances.

Because of rapidly growing subscriber populations, advances in cellular communication technology, increasingly capable user terminals, and the expanding range of mobile applications, cellular networks have experienced a significant increase in data traffic, the dominant part of which is carried by the http protocol. Understanding the characteristics of this traffic is important for network design, traffic modeling, resource planning and network control. In this study we present a comprehensive characterization study of mobile http-based traffic using packet level traces collected in a large cellular network. We analyze the traffic using metrics at packet level, flow level and session level. For each metric, we conduct a comparison between traffic from different applications, as well as comparison to traffic in a wired network. Finally, we discuss the implications of our findings for better resource utilization in cellular infrastructures.

Data traffic in cellular networks increased tremendously over the past few years and this growth is predicted to continue over the next few years. Due to differences in access technology and user behavior, the characteristics of cellular traffic can differ from existing results for wireline traffic. In this study we focus on understanding the flow rates and on the relationship between the rates and other flow properties by analyzing packet level traces collected in a large cellular network. To understand the limiting factors of the flow rates, we further analyze the underlying causes behind the observed rates, e.g.,network congestion, access link or end host configuration. Our study extends other related work by conducting the analysis from a unique dimension, the comparison with traffic in wired networks, to reveal the unique properties of cellular traffic. We find that they differ in variability and in the dominant rate limiting factors.