An information-theoretic approach to queuing in wireless channels with large delay bounds

Abstract

Abstract-Queuing theory allows the design of communication links that can provision for Quality of Service (QoS) in time-varying channels, such as mobile wireless channels, by considering an idealized queuing system that abstracts out the physical layer. Similarly, information theory allows the design and analysis of channel codes that can guarantee low decoding error probability by adapting to the time-varying channel. Whereas other researchers have attempted cross-layer design methods that combine these two approaches, these have been limited to specific choices of practical channel codes. There have been few attempts to combine these two theories to specify the ultimate limit of delay-constrained communications. This paper presents an approach that obtains a QoS exponent, by combining the queuing and information theoretic models; in particular, by considering information-theoretically optimal channel codes. Calculations show that such a joint approach yields substantial improvement in QoS performance in a variety of communication scenarios.

High-SNR Analysis of Outage-Limited Communications With Bursty and Delay-Limited Information

“…In [5], Rajan et al derive optimal delay-bounded schedulers for transmission of constant-rate traffic over finitestate fading channels. In [6], Negi and Goel apply the effective capacity [9] and error exponent techniques to find the coderate allocation that maximizes the decay rate of the asymptotic probability of error for a given asymptotically-large delay requirement. Similar to [3] and [4], the proposed dynamic code-rate allocation in [6] is in response to the current channel fading and is possible by assuming CSI knowledge at the transmitter.…”

“…Remark 1: The condition that each bit be transmitted over all timeslots in the coding block 5 , together with the firstcome first-transmit service discipline, makes it so that every T timeslots, the RT oldest bits are instantaneously removed 6 from the queue and are transmitted over the next T timeslots. We assume that it is only at the end of the T timeslots that all the RT bits are decoded by the decoder.…”

“…5 Currently, all known minimum-delay DMT optimal codes over any fading channel with non-zero coefficients ask that each bit be transmitted over each timeslot. 6 If an insufficient number of bits exists in the buffer, null bits are used and the rate is maintained. It is easy to show that, in the asymptotic scale of interest, the use of null-bits does not incur any change in the SNR exponent of the probability of error.…”

Optimal Operating Point in MIMO Channel for Delay-Sensitive and Bursty Traffic

“…Our goal is to find the optimal operating point on the diversity-multiplexing region that minimizes the end-to-end bit loss probability due to both the delay bound violation and the MIMO channel decoding errors. Our work, in spirit, is related to [3]. In [3], the authors study the optimal (time-varying) encoding rate of a system consisting of a delay-sensitive constant rate source over a time-varying channel.…”

“…Our work, in spirit, is related to [3]. In [3], the authors study the optimal (time-varying) encoding rate of a system consisting of a delay-sensitive constant rate source over a time-varying channel. The channel code is not assumed ideal.…”