We present cluster-based publish/subscribe, a novel architecture that is not only resilient to event broker failures but also provides load balancing and fast event dissemination service. Our proposed approach achieves fault tolerance by organizing event brokers in clusters. Multiple intercluster links provide continuous availability of dissemination service in presence of broker failure without requiring subscription retransmission or reconstruction of broker overlay. Furthermore, the proposed architecture provides a fast event dissemination infrastructure that significantly reduces subscription and publication dissemination traffic and load on event brokers. Our experimental results show that even in the presence of 10% failure rate in broker network, event dissemination is not interrupted and dissemination speed and load are not affected significantly.