Design a cloud-native historical data processing system: data in CSV, Avro, PDF; accessed by Dataproc, BigQuery, Compute Engine; daily batch ingestion; performance not critical; maximize availability. How should you store the data?

Storing data in a multi-regional Cloud Storage bucket is the best choice because it maximizes availability, a key requirement. Cloud Storage is a highly durable and available object storage service, and multi-regional buckets offer the highest level of availability and geographic redundancy. It also supports all specified data formats (CSV, Avro, PDF) and allows direct access from Dataproc, BigQuery, and Compute Engine without needing to move or copy data, simplifying the architecture. A regional Cloud Storage bucket offers high availability but not the maximum availability of a multi-regional one. Storing data in BigQuery is not ideal for PDF files and can be costly for raw, diverse formats if not immediately structured for analytics. A Dataproc cluster with HDFS is not cloud-native, introduces operational overhead, and doesn't provide the same level of availability or direct integration with BigQuery and Compute Engine for diverse access patterns.