1 A Joint Congestion Control, Routing, and Scheduling Algorithm in Multihop Wireless Networks with Heterogeneous Flows Phuong Luu Vo, Nguyen H. Tran, Choong Seon Hong, *KiJoon Chae Kyung Hee University, *Ewha Womans University, Korea {phuongvo, nguyenth, cshong}@khu.ac.kr, *kjchae@ewha.ac.kr Abstract—We consider the network with two kinds of traffic: inelastic and elastic traffic. The inelastic traffic requires fixed throughput, high priority while the elastic traffic has controllable rate and low priority. Giving the fixed rate of inelastic traffic, how to inject the elastic traffic into the network to achieve the maximum utility of elastic traffic is solved in this paper. The Lagrangian Duality method is applied to solve the op- timization problem. We decompose the Lagrangian into sub- problems, and each sub-problem associates with each layer. The convexity of the primal problem guarantees the duality gap between primal and dual solutions is zero. The Lagrange multipliers, which are indeed the queue length on nodes for every destinations, implicitly update according to subgradient algorithm. The joint algorithm for rate control, routing, and scheduling is proposed. However, the scheduling is Max-weight scheduling and centralized algorithm actually. The Greedy dis- tributed scheduling is introduced to implement scheduling in a distributed sense. I. I NTRODUCTION Network optimization and control is an active research area [1]–[4]. There are two main kinds of network formu- lations: node-centric and link-centric [3]. The node-centric formulation maximizes the utility such that all queues are stable: total transmitting rate and incoming flow of a queue must less than the outgoing flow. On the other hand, the link- centric formulation uses the capacity constraint: total of load of all flows on a link must be less than the capacity of the link. Whereas link-centric formulation need a routing matrix in the formulation, the node-centric does not; therefore, the dynamic routing is also solved in the node-centric formulation. Recent papers apply the frameworks for networks with heterogeneous flows: elastic and inelastic flows [5]–[8]. By using the link-centric formulation, giving the inelastic rate, the authors in [6] maximize the utility of elastic traffic while load- balancing the inelastic traffic on some predefine routes. Also using the link-centric formulation, the optimization problem is solved in [7] with the additional constraint: the probability of missing the deadline packets less than a threshold. All these papers use link-centric formulation, therefore, the routing matrix must be a priori. This work was supported by the IT R&D program of MKE/KEIT [10035245: Study on Architecture of Future Internet to Support Mobile Environments and Network Diversity]. Dr. CS Hong is the corresponding author. Our paper applies the node-centric formulation to solve the problem, so the dynamic routing is integrated naturally. We don’t cover the end-to-end delay constraint in the scope, but the priority of inelastic traffic in using the links in wireless environment is considered. The rate of inelastic traffic is fixed (the source demand). We want to inject the elastic traffic such that maximizing the utility of elastic traffic while all the queues in the network keep stable. Our contributions in this paper are: 1) Applying the node-centric formulation in the cross- layer design for the multihop wireless network with the requirement of fixed the inelastic rate (source demand) and optimal utility of elastic traffic. 2) Proposing the rate control for elastic traffic, the dis- tributed routing and scheduling for both kinds of traffic. 3) Providing the simulation of the impact of higher priority of inelastic rate demand on the lower priority elastic traffic. II. PROBLEM FORMULATION The network is modeled by a directed graph =(, ), where is the set of nodes, and is the set of links. The network has two kinds of flows: inelastic flows and elastic flows. Define as the set of inelastic flows. Each inelastic flow maps a pair of two nodes: source node and destination node. Let and be the sets of inelastic source nodes and destination nodes respectively. We have ⊂ and ⊂ . Set is the rate of flow . just depends on the demand of multimedia service, and it is a constant. We assume that the inelastic rate is always admissible by the network. Define is the set of elastic flows. Each elastic flow maps a pair of two nodes: source node and destination node. Define and as the sets of elastic source nodes and destination nodes respectively. We also have ⊂ , . In this paper, we assume that the sets of of destination nodes of inelastic and elastic flows are disjoint. Each elastic flow is associated with a utility function (.), which is concave, twice-differentiable, and non-decreasing. For example, log() where is the rate of the flow is a utility function. is the rate of flow , and x () is the rate vector of all elastic flows. Here we want to control the rate to archive the maximum total of utility. Link-Rate Region 347 978-1-61284-663-7/11/$26.00 ©2011 IEEE ICOIN 2011