This paper proposes a method to analyze the delay distribution at the transmitter side for cooperative diversity wireless networks with link adaptation, which uses decode-and-forward as a relay strategy. The equivalent end-to-end channel model is captured by a finite state Markov channel model representing the multiple states of the Nakagami fading channel, and the delay statistics are obtained by using matrix geometry method. Cooperative diversity wireless network has a better delay performance compared with noncooperation wireless network.Adaptive modulation and coding (AMC) is used in most of the wireless networks to increase the transmission rate by exploiting the wireless channel variations [8][9][10][11]. The problem of applying AMC to a wireless cooperative relay network was investigated in [12]. In [12], a discrete rate and power adaptation policy was proposed for the fixed-DF relay channel. Transmission during outage was counted as the main cause of error, because capacity achieving codes were assumed, which resolve errors because of channel noise. Although the design of strong and reliable error correction codes has played a key role in error-control for applications with strict delay requirements (e.g., voice), the deployment of delay-tolerant data services in wireless networks makes automatic repeat request (ARQ)-based error-control very attractive to counteract the residual errors without using costly error correction codes at the physical layer.Automatic repeat request protocol is an error-control method of the link layer, which guarantees reliable transferring of packets by retransmitting packets after negative acknowledgments (NACKs) are reported on the feedback channel. There are three basic types of ARQ protocols: stop-and-wait (SW) ARQ, go-back-N (GBN) ARQ, and selective-repeat (SR) ARQ protocols. SR-ARQ is the most efficient one (in terms of throughput); however, GBN-ARQ is superior to the SR-ARQ protocol in terms of implementation simplicity. In [13], an approximate expression of the average delay for SR-ARQ was obtained for a two-state Markov channel under heavy traffic condition. In a multirate wireless network, the delay statistics for SR-ARQ with noninstantaneous feedback delay was derived in [14]. Note that the availability of radio link-level delay statistics allows wireless network design and engineering under statistical delay constraints of the form P r¹delay > D max º < P t instead of those based on the average delay or delay bounds.A finite-state Markov channel (FSMC) model was proposed for Rayleigh and the more general Nakagami-m fading channels in the literature [15][16][17][18][19]. In [15], Zheng et al. presented the crosslayer analytical framework in analyzing the QoS of the DF relaying wireless networks, where AMC is employed at the physical layer under the conditions of unsaturated traffic and finite-length queue at the data link layer. On considering the characteristics of DF relaying protocol at the physical layer, the authors in [15] first proposed modeling a two-hop DF r...