This thesis explores strategies for resource allocation in the wireless channel under time constraints. These are considered in two different manifestations, leading to a two-part structure of the thesis. 1) The first part explores delay-constrained rate requests in multiuser downlink scenarios. Although this topic has been explored repeatedly in the past, there is a lack of approaches that are practical from a computational complexity perspective, and flexible enough to adapt to a wide variety of channel conditions, rate constraints and delay limitations. The first contribution of this thesis is therefore the discussion of properties of widespread resource allocation strategies with respect to their suitability for delay-constrained transmission of data. Current implementations of scheduling algorithms generally lack information about rate demands and delay requirements of higher layers in the OSI model. As long as the channel quality does not change too quickly, applications typically adapt their demands to the current possibilities the communication system provides. This is, however, also a strategy to compensate for the inability of the scheduler to serve all requests in time due to the lack of commonly implemented information exchange between the application layer and the scheduler. Therefore, the very different limitations regarding required rate and acceptable delay for normal web-browsing or video streaming cannot be considered in resource allocation. This thesis therefore suggests a new concept - Power-Controlled Cross-Layer Scheduling (PCCLS) - that exploits exactly this information, opening up a range of possibilities that are not available in today's standard implementations. Compared to proportional fair (PF) scheduling, PCCLS adds a control loop that compensates for slow changes in channel attenuation, thus providing a constant long-term channel quality that makes the achievement of specific rate- and delay- requirements possible. The faster block-fading is not compensated for, i.e. it can be taken advantage of. In addition, PCCLS excludes certain users from scheduling as soon as the required rate has been achieved in a given delay-window. This is done even when the transmit buffer still provides data to be transmitted, freeing up resources for other users. Through numerical simulations, PCCLS is shown to achieve the given rate- and delay- constraints for a general scenario based on realistic requirements. Proportional fair scheduling, on the other hand, is demonstrated to spend too much energy on users with low rate requests, which leads to a failure regarding the requested rates of the more demanding users. The second main contribution in this first part of the thesis derives the probability of an outage for resource allocation schemes that are based on a memoryless selection of the user with the largest scheduling metric. Well-known members of this species of scheduling schemes are the proportional fair- and the opportunistic scheduler. The analysis does not constrain the number of users, their respective channel (fading) statistics or rate demands. The general derivation is subsequently specialized to the original implementations of proportional fair and opportunistic scheduling; simulations demonstrate the perfect match between analytic and numeric approaches. Finally, it is shown that accepting more relaxed delay requirements results in a huge reduction of the experienced outage probability. 2) While the first part of this thesis explores strategies for efficient use of resources when there are user-imposed time limitations in the form of rate- and delay- constraints, the second part considers time limitations that originate from the channel. These are experienced in the form of periods where a transmission is economically feasible, and intensified from an algorithmical point of view when the channel is not ergodic during these windows of opportunity. In practice, these scenarios can be observed in ad-hoc communications, specifically between vehicles on opposing highway lanes or in body-area communications. Sum-power constraints directly lead to a waterfilling solution, raising the question for the optimum transmit decision threshold (i.e., the waterlevel) when the future evolution of the channel is unknown. A new approach to directly calculate this decision threshold is derived, and its properties are explored analytically. Subsequently, the gathered insights are used to compare acausal, causal and causal-adaptive implementations regarding efficient use of the limited transmit power budget. The second part of this thesis concludes with the conceptual discussion of a strategy for recognizing transmit opportunities when the channel is available during recurring time windows.