Lazaros F. Merakos

Performance analysis of reservation random access protocols for wireless access networks

A key issue in future wireless access networks is the design of multiaccess protocols that enable dispersed voice and data terminals to efficiently share the terminal-to-base radio channel. A promising class of multiaccess protocols for wireless access is that of reservation random access (RRA) protocols, in which voice talkspurt packets and data packets are statistically multiplexed over TDMA frames. Voice terminals obtain reservations for their talkspurts using a random access protocol. An analysis method for evaluating the performance of a class of RRA protocols is developed. The method is based on a Markovian model, whose stationary solution is obtained via an iterative computational procedure based on matrix decomposition techniques. Subject to a maximum delay constraint, the voice packet dropping rate and distribution of the number of packets dropped from a talkspurt are evaluated for three RRA protocol examples, and comparisons are made with the performance of an ideal hypothetical multiplexer. The transient performance of RRA protocols is evaluated by simulating their response to sudden changes in the number of active voice terminals. Finally, suggestions for the design of integrated voice/data RRA protocols are made. >

Voice/data channel access integration in TDMA digital cellular networks

The integration of digital data capabilities in the soon to be widely deployed digital cellular networks, which were primarily designed for voice communications, offers a low-cost way to capture the large and ever growing market for mobile data services. The authors propose and evaluate a multiaccess protocol for integrating data traffic in the E(nhanced)-TDMA voice system with digital speech interpolation, which is an enhancement of the emerging North American digital cellular standard. The proposed, protocol combines random access with slot reservation mechanisms to statistically multiplex data packets with speech spurt packets over the shared terminal-to-base air channel. The integrated protocol requires no modification in the voice access protocol used in the E-TDMA system, and can attain performance close to that of an ideal voice/data multiplexer. Furthermore, the protocol may enable multislot assignment per TDMA frame to match the throughput needs of individual data terminals, and can accommodate application-dependent data transmission priorities. >

Delay analysis of the n-ary stack random-access algorithm

An analysis is presented of the delay performance of the n-ary stack algorithm, an easy-to-implement random-access algorithm that uses collision-no-collision binary feedback and requires only limited feedback sensing by the users. Utilizing the regenerative character of the stochastic processes that are associated with the random-access system, tight upper and lower bounds on the mean packet delay and on the packet delay variance are derived. It is shown that the algorithm with n=3 has the best delay characteristics. Simulation results for the percentiles of the packet delay distribution are given. >

Interconnection of CSMA Local Area Networks: The Frequency Division Approach

This paper considers the interconnection of two local area networks, each of which connects a large population of users via a carrier-sense multiple-access (CSMA) channel. In each network a bridge node receives internetwork packets from the local users and forwards them to the bridge node of the other network via a point-to-point link; the bridge node of the destination network queues these internetwork packets for subsequent broadcasting to the local users. The available broadcast channel in each of the two networks is frequency divided into a bridge node subchannel, used for bridge node transmissions, and a CSMA user subchannel, used for user transmissions. For the interconnected system, stability conditions, optimal bandwidth allocations, and throughput delay characteristics are found. Also it is shown that the performance of a single CSMA network can be substantially improved by partitioning it into smaller interconnected networks.

A reservation protocol for packet voice and data integration in unidirectional bus networks

A movable boundary protocol is proposed for integrating packet voice and data in unidirectional bus networks. The head station on the bus learns the number of ready-to-transmit voice stations by reading a request bit in the header of the received packets and allocates the exact number of voice slots needed in each frame. The protocol guarantees that the maximum delay to transmit a voice packet will be less than the round-trip propagation delay at the head station plus twice the time needed to form the packet. The average data packet delay is evaluated via approximate analysis and simulation, for the case in which the voice-reserved slots in a frame are contiguous and for the case in which they are evenly distributed. >

Interconnection of CSMA/CD LANs via an N-port bridge

The interconnection of CSMA/CD LANs using an N-port bridge is considered. The stability and throughput-delay performance of the interconnected system is analyzed approximately by decomposing it into N consistent but otherwise independent LAN subsystems. The performance of a LAN subsystem with bridge priority is evaluated, and the results are used for the performance evaluation of the interconnected system. Simulation results indicate that the decomposition approach provides reasonably accurate performance prediction. The bridge interconnection system is compared to a system in which the bridge is replaced by a repeater, and the performance advantages of the former system over the latter are quantified.<<ETX>>

Interconnection of high speed token ring LANs: a switch-connection approach

The use of a switch for the interconnection of two high-speed token-ring LANs to avoid congestion at the bridge caused by the large bridge processing time is proposed. A simple network access method is described, and the delay performance of the switch-connected system is analyzed and compared with that of the bridge-connected system. For high-speed token-ring LANs, the comparison shows that the bridge will be a bottleneck unless its processing time is significantly reduced. The proposed interconnection using the switch eliminates the bottleneck problem and, depending on the bridge processing time, it can attain better delay performance than the bridge-connected system over a wide range of system parameters.<<ETX>>

Performance analysis of reservation random access protocols for cellular packet communications

An analysis method for evaluating the performance of a class of RRA (reservation random access) protocols is developed. The method is based on a Markovian model whose stationary solution is obtained via an iterative computational procedure based on matrix decomposition techniques. Subject to a maximum delay constraint, the steady state probability of voice packet dropping is evaluated for three RRA protocol examples and comparisons are made with the performance of an ideal hypothetical multiplexer. Suggestions for the design of integrated voice/data RRA protocols are made. The transient performance of RRA protocols is evaluated.<<ETX>>

Analysis of a priority leaky bucket regulator for virtual path traffic in ATM networks

The proposed priority traffic regulator maintains a separate cell buffer for each traffic priority to be able to trade off cell delay and loss (or tagging) on the basis of the quality of service requirements for each priority class. A versatile Markovian point process model is adopted for both priority and low-priority sources. The joint buffer content distribution and cell loss probabilities of the two priority classes are obtained. The obtained results capture the effect of the shaper/policer and source parameters on the cell loss/delay performance, and may be used for design optimization.

Shaping of virtual path traffic for ATM B-ISDN

Two mechanisms for shaping private network traffic in asynchronous transfer mode (ATM) broadband ISDN (B-ISDN) are proposed and analyzed. The shapers utilize the information about the network policer and implement a form of priority service to assure satisfactory levels of cell loss and delay for traffic streams with differing service requirements multiplexed into a policed ATM virtual path. The performance of the shapers is analyzed using stochastic-fluid methods and Markov-modulated models of bursty traffic sources. The numerical results illustrate the performance benefits and flexibility obtainable through the use of the proposed shapers in comparison to a simple policed path without traffic shaping.<<ETX>>