Capacity of Large-Scale CSMA Wireless Networks

Abstract

In the literature, asymptotic studies of multi-hop wireless network capacity often consider only centralized and deterministic TDMA (time-division multi-access) coordination schemes. There have been fewer studies of the asymptotic capacity of large-scale wireless networks based on CSMA (carriersensing multi-access), which schedules transmissions in a distributed and random manner. With the rapid and widespread adoption of CSMA technology, a critical question is that whether CSMA networks can be as scalable as TDMA networks. To answer this question and explore the capacity of CSMA networks, we first formulate the models of CSMA protocols to take into account the unique CSMA characteristics not captured by existing interference models in the literature. These CSMA models determine the feasible states, and consequently the capacity of CSMA networks. We then study the throughput efficiency of CSMA scheduling as compared to TDMA. Finally, we tune the CSMA parameters so as to maximize the throughput to the optimal order. As a result, we show that CSMA can achieve throughput as Ω( 1 √ n ), the same order as optimal centralized TDMA, on uniform random networks. Our CSMA scheme makes use of an efficient backbone-peripheral routing scheme and a careful design of dual carrier-sensing and dual channel scheme. We also address the implementation issues of our CSMA scheme.

The Capacity of Wireless CSMA/CA Networks

“…Let be the overall average interarrival time at including its frozen time. Assuming stability and following the same logic as before, the fraction of time that transmits must then be (23) where the second equality holds from (21). From Theorem 1, we know that the relation is one-to-one, and thus, from (21) and (23), and must be unique for each .…”

“…Assuming stability, no packets accumulate in the queues, and the total time that transmits in is approximately , since each packet is transmitted times on average. The fraction of time that transmits in the unfrozen timeline is then (21) where the right-hand side is the fraction of time that transmits when it is alone, from (11). Isolating in (21) ).…”

Forensic Analysis of Packet Losses in Wireless Networks

“…α is the path loss exponent. The collision model: There are several models used to capture collisions in the literature [19]. We use the SINR (Signalto-Interference-and-Noise) physical model, where node v j successfully receives the transmission from node v i iff :…”

General capacity scaling of wireless networks

“…N0+ <i,j>∈K t /<u,v> P · (xi−xv|) ; x u denotes the position of node u, |x u − x v | represents the Euclidean distance between node u and node v; (·) denotes the power attenuation function that is assumed to depends only on the distance between the transmitter and receiver [1]- [3], [15] …”