TY - JOUR

T1 - Performance analysis of differentiated resource-sharing in a wireless ad-hoc network

AU - Roijers, Frank

AU - van den Berg, Hans

AU - Mandjes, Michel

PY - 2010/7

Y1 - 2010/7

N2 - In this paper we model and analyze a relay node in a wireless ad-hoc network; the capacity available at this node is used to both transmit traffic from the source nodes (towards the relay node), and to serve traffic at the relay node (so that it can be forwarded to successor nodes). Clearly, a network node that is used more heavily than others is prone to becoming a performance bottleneck. Therefore we consider the situation that the relay node obtains a share of the capacity that is m times as large as the share that each source node receives. The main performance metrics considered are the workload at the relay node and the average overall flow transfer time, i.e., the average time required to transmit a flow from a source node via the relay node to the destination. Our aim is to find expressions for these performance metrics for a general resource-sharing ratio m, as well as a general flow-size distribution. The analysis consists of the following steps. First, for the special case of exponential flow sizes we analyze the source-node dynamics, as well as the workload at the relay node by a fluid-flow queueing model. Then we observe from extensive numerical experimentation over a broad set of parameter values that the distribution of the number of active source nodes is actually insensitive to the flow-size distribution. Using this remarkable (empirical) result as an approximation assumption, we obtain explicit expressions for both the mean workload at the relay node and the overall flow transfer time, both for general flow-size distributions.

AB - In this paper we model and analyze a relay node in a wireless ad-hoc network; the capacity available at this node is used to both transmit traffic from the source nodes (towards the relay node), and to serve traffic at the relay node (so that it can be forwarded to successor nodes). Clearly, a network node that is used more heavily than others is prone to becoming a performance bottleneck. Therefore we consider the situation that the relay node obtains a share of the capacity that is m times as large as the share that each source node receives. The main performance metrics considered are the workload at the relay node and the average overall flow transfer time, i.e., the average time required to transmit a flow from a source node via the relay node to the destination. Our aim is to find expressions for these performance metrics for a general resource-sharing ratio m, as well as a general flow-size distribution. The analysis consists of the following steps. First, for the special case of exponential flow sizes we analyze the source-node dynamics, as well as the workload at the relay node by a fluid-flow queueing model. Then we observe from extensive numerical experimentation over a broad set of parameter values that the distribution of the number of active source nodes is actually insensitive to the flow-size distribution. Using this remarkable (empirical) result as an approximation assumption, we obtain explicit expressions for both the mean workload at the relay node and the overall flow transfer time, both for general flow-size distributions.

KW - Bottleneck

KW - Feedback fluid-queue

KW - Insensitivity

KW - Resource sharing

KW - Transfer time

UR - http://www.scopus.com/inward/record.url?scp=77955309288&partnerID=8YFLogxK

U2 - 10.1016/j.peva.2010.01.005

DO - 10.1016/j.peva.2010.01.005

M3 - Article

AN - SCOPUS:77955309288

SN - 0166-5316

VL - 67

SP - 528

EP - 547

JO - Performance evaluation

JF - Performance evaluation

IS - 7

ER -