This dissertation considers the design, evaluation and optimization of algorithms/ techniques/ system parameters for distributed wireless networks specifically ad-hoc and cognitive wireless networks. In the first part of the dissertation, we consider ad-hoc networks using opportunistic carrier sense multiple access (CSMA) protocols. The key challenge in optimizing the performance of such systems is to find a good compromise among three interdependent quantities: the density and channel quality of the scheduled transmitters, and the resulting interference seen at receivers. We propose two new channel-aware slotted CSMA protocols and study the tradeoffs they achieve amongst these quantities. In particular, we show that when properly optimized these protocols offer substantial improvements relative to regular CSMA -- particularly when the density of nodes is moderate to high. Moreover, we show that a simple quantile based opportunistic CSMA protocol can achieve robust performance gains without requiring careful parameter optimization. In the second part of the dissertation, we study a cognitive wireless network where licensed (primary) users and unlicensed 'cognitive' (secondary) users coexist on shared spectrum. In this context, many system design parameters affect the joint performance, e.g., outage and capacity, seen by the two user types. We explore the performance dependencies between primary and secondary users from a spatial reuse perspective, in particular, in terms of the outage probability, node density and joint network capacity. From the design perspective the key system parameters determining the joint transmission capacity, and tradeoffs, are the detection radius (detection signal to interference and noise power ratio (SINR) threshold) and decoding SINR threshold. We show how the joint network capacity region can be optimized by varying these parameters. In the third part of the dissertation, we consider a cognitive network in a heterogeneous environment, including indoor and outdoor transmissions. We characterize the joint network capacity region under three different spectrum (white space) detection techniques which have different degrees of radio frequency (RF) - environment awareness. We show that cognitive devices relying only on the classical signal energy detection method perform poorly due to limitations on detecting primary transmitters in environments with indoor shadowing. This can be circumvented through direct use (e.g., database access) of location information on primary transmitters, or better yet, on that of primary receivers. We also show that if cognitive devices have positioning information, then the secondary network's capacity increases monotonically with increased indoor shadowing in the environment. This dissertation extends the recent efforts in using stochastic geometric models to capture large scale performance characteristics of wireless systems. It demonstrates the usefulness of these models towards understanding the impact of physical /medium access (MAC) layer parameters and how they might be optimized. / text