In this third installment of a series on high availability (HA) and disaster recovery (DR) with Redpanda, the focus is on deploying Redpanda clusters across multiple availability zones (multi-AZ) to enhance system resilience. The post explains that by distributing brokers over three or more zones, the architecture minimizes single points of failure, as the Raft consensus algorithm ensures that the cluster can still function if one zone fails. While a multi-AZ setup provides increased availability, it also incurs higher costs and potential latency due to the network distance between zones and additional data transfer charges. The introduction of follower fetching, a feature in Redpanda 23.2, allows consumers to read data from the closest replica, which reduces cross-zone traffic and associated costs. The article discusses deployment options using Redpanda Cloud or self-hosted solutions with Terraform and Ansible, emphasizing the balance between cost, complexity, and the need for extreme availability, with a promise of exploring multi-region deployments in the next post.