Roger Kalden

Wireless Internet access based on GPRS

Enabling wireless Internet access is one of the upcoming challenges for mobile radio network operators. The General Packet Radio Service is the packet-switched extension of GSM and was developed to facilitate access to IP-based services compared to existing circuit-switched services provided by GSM. Besides an overview on the basic concept, network architecture, and protocols of GPRS, this article discusses the performance the end user perceives when retrieving information from the Web using this access technology. This discussion is based on results obtained from a protocol simulator comprising not only the radio interface protocols, but also the relevant Internet protocols as well as a characteristic application model. The results show that GPRS provides bandwidth-efficient support for bursty applications like Web access.

Searching for Self-Similarity in GPRS

Based on measurements in live GPRS networks, the degree of self-similarity for the aggregated WAP and WEB traffic is investigated by utilizing six well established Hurst parameter estimators. We show that in particular WAP traffic is long-range dependent and its scaling for time scales below the average page duration is not second order self similar. WAP over UDP can also determine the overall traffic scaling, if it is the majority traffic. Finally we observe that the minor traffic exhibits a larger Hurst value than the aggregated traffic, in case of WAP as well as in case of WEB traffic.

Wireless service usage and traffic characteristics in GPRS networks

With the deployment of packet-switched wireless GPRS access networks, new wireless services are introduced, such as WAP and MMS, and existing Internet applications are put into a new mobile usage context. Optimal design and dimensioning of GPRS requires the knowledge of subscribers usage of different services and their traffic characteristics. For this purpose, we have investigated GPRS usage and traffic patterns based on measurements taken in two live GPRS networks. The networks, revealing very different application mixtures, are investigated in terms of session duration and data volume, packet size distribution, application usage frequency and data volume, as well as user mobility.

Searching for mobile mice and elephants in GPRS networks

In this paper, we provide statistics on large-scale traffic measurements that have been made in a live cellular GPRS network. We show that the current GPRS traffic is dominated by HTTP- and WAP-based applications, and further show in what direction (uplink and/or downlink) the ensuing traffic is transmitted as well as statistics on the length of the flows on a per application basis. We find that the results differ compared to those found in similar measurement studies that have been carried out in the fixed Internet. In particular, our results suggest that flows are shorter in cellular networks, and that the so-called mice and elephant effect, whereby a small part of the flows make up a majority of the bytes transferred, is not prevalent in GPRS networks. Instead, extremely short flows account for the majority of the total flows and bytes transferred. We also discuss the impact of this finding on future research on TCP loss recovery.

Measuring and Modeling of Application Flow Length in Commercial GPRS Networks

New mobile access networks provide reasonable high bandwidth to allow true Internet access. This paper models two dominant applications of those networks. One application, WAP, is novel and specific to mobile networks, the other is HTTP, which is already dominantly present in the Internet. However, our measurements reveal that WAP traffic is the dominant application in mobile networks. Therefore, this paper models in particular the flow length of WAP and as comparison the flow length of HTTP. The flow length is a good approximation of the object size structure and hence can be used in further simulation or analytical modeling of mobile networks. We apply different distribution fitting techniques and the KS goodness of fit test to specify exactly the underlying distribution of the empirical data