Comparison of Storage Architectures for Data Warehousing

This technical report provides a detailed comparison of two storage architectures used in modern data warehouse platforms: Compute Adjacent Storage Architecture (CASA) and Remote Object Storage Architecture. The report outlines the performance and cost differentials between these two architectures, specifically when configured to handle 2.5PB of data. It begins by defining remote storage and its reliance on spinning disk drives connected via a network, contrasting this with CASA, which utilizes NVMe SSDs directly attached to processors. The document presents a cost breakdown over a five-year period, including capital and operating expenses, and examines the performance at the IO layer for both architectures. Additionally, it discusses the implications of declining SSD costs and the shift towards NVMe technology, emphasizing the efficiency and performance gains achievable with CASA. The report concludes with a comparison of hardware requirements and the overall effectiveness of each architecture in data processing and storage management.