The rapid growth of data volume and analytical complexity has exposed the scalability and performance limits of traditional relational data warehouses (Rel-DWs). In response, column-family NoSQL systems such as HBase and Cassandra have emerged as viable foundations for Big Data analytics and OLAP workloads. However, designing an efficient Column-Family NoSQL Data Warehouse (CN-DW) from relational schemas remains challenging, particularly with respect to attribute grouping, redundancy management, and workload adaptability. This paper presents a critical survey of CN-DW design and optimization approaches, classifying existing methods into naive, advanced, and optimized categories, as well as normalized and denormalized schema strategies. Special attention is given to workload-aware optimization based on unsupervised clustering and metaheuristics, including K-means, K-medoids with PSO, and CLARANS. The survey highlights their strengths and limitations and outlines emerging directions toward adaptive and self-optimizing CN-DW architectures.