Portal de Periódicos da CAPES

interview

2013; Institution of Engineering and Technology; Volume: 49; Issue: 12 Linguagem: Inglês

10.1049/el.2013.1735

1350-911X

John N. Daigle,

Professor John Daigle from the University of Mississippi in the US talks about the research behind his Letter ‘Dimensioning of dual leaky bucket parameters for regulating voice sources’ on page 756. My interests are mainly in two areas: modelling and analysis of mammalian inter and intracellular communications, and information transfer in the future Internet. In the former area we are teamed with biologists and medical professionals, and our objectives are in the direction of characterising the response of cells to external stimuli. In the latter area, we are interested in developing protocols that provide reliable and efficient data delivery in a multi-source, multi-destination setting over channels with heterogeneous capacities, delay and loss characteristics. We are specifically interested in applications of fountain codes and their recent implementations, including Luby transform and raptor codes. Professor John Daigle The motivation is the importance of the problem. As the quality of transmission is improved and higher capacity transmission systems continue to be deployed, it is only a matter of time until all communication services are packet-based. The transition from circuit switched to packet-based communication in the case of voice has a number of interesting challenges. On the one hand, it is important to assure that adequate resources are allocated to each call, but on the other hand it would be prohibitively expensive to try to keep track of every call and allocate resources on a call-by-call basis. Thus, it makes sense to develop mechanisms within that manage resource for large numbers of VoIP sessions without having to consider performance of individual sessions. Since communication channels, especially in the backbone, can handle a very large number of VoIP calls, it is reasonable to expect that groups of calls would have performance close to the average. If, in addition, the individual calls in the large group have more or less the same capacity demand, then it would be easy to track of the number of calls being carried over short periods of time by simply tracking the total packet transmission rate. One way to push VoIP sessions in the direction of equal capacity demand is to shape the traffic at the network ingress. The subject of the work presented here is how to shape the traffic to achieve a prescribed quality of service. The difference is primarily in the way we view the system. We view the speaker as operating in a sequence of cycles, where a cycle is defined as the point at which a speaker begins a talk spurt. At that moment, the speaker finds a number of tokens present in a token bucket, and for each token present, the speaker can send one voice packet into the system. If there are insufficient tokens present at the beginning of a cycle, then some voice packets are dropped. The number of tokens present at the beginning of a cycle is a discrete parameter, discrete state Markov chain, and it is reasonably straightforward to solve its equilibrium probabilities. The packet loss rate conditioned on the number of tokens present at the beginning of a cycle is easy to calculate, and thus the loss behaviour over a typical cycle is easy to compute. The approach has the advantage of avoiding the details of the time-dependent operation over the cycle, which has no bearing whatsoever on the results. In addition, the approach is robust with respect to the form of the distributions of the talk and silent periods. We are deeply involved in experimentation to quantify the movement of specific molecules within a mammalian cell. We have found that Markov chains provide an excellent framework for characterising molecule production rates and response times to stimuli. In our information transfer thread, we are implementing many-to-one file transfer protocols based on Raptor codes in Raspberry Pi computers and measuring performance over specially configured local area networks. We also have a number of ongoing smaller efforts in the area of media access protocols for ad hoc wireless networks and localisation of sensors in wireless networks where there is a significant fading component.