In this thesis we present the top-k/w matching model, a novel publish/subscribe matching model which enables a subscriber to control the number of publications it will receive per subscription within a predefined time period. In this model, each subscription defines an arbitrary and time-independent scoring function and parameters k and window-size w such that, at a point in time t, parameter k restricts the number of delivered publications to the k best scored publications that are published between t-w and t. We propose novel algorithms for efficient top-k/w processing in environments publishing data objects at high rates. Existing solutions are typically tailored to specific scoring functions, and thus we define a generic data stream processing model that is independent of data representation and scoring functions. Our complexity analysis and extensive comparative evaluation show that our algorithms significantly outperform the competing approaches in both memory consumption and processing efficiency. For each of the commonly used routing strategies in both centralized and distributed environments we explain how it can be adapted to support the proposed top-k/w matching model. We identify and evaluate routing strategies which are particularly well suited for large-scale top-k/w publish/subscribe systems. Additionally, we present D-ZaLaPS, a distributed top-k/w publish/subscribe system based on a peer-to-peer structured overlay. The experimental evaluation shows that D-ZaLaPS is scalable both for increasing number of peers and subscriptions. Finally, using the case study with k-nearest neighbor subscriptions, we experimentally evaluate and compare the top-k/w and Boolean matching models. This experimental evaluation shows that, assuming the subscribers are interested in top-k publications in the sliding window, the top-k/w matching model can significantly reduce message overhead in the system when compared to the Boolean matching model, and this, together with built-in support for flexible subscriptions and control of the number of matching publications in the top-k/w model, represents a significant enhancement of publish/subscribe systems.