1: Snowflake AI Data Cloud Features and Architecture

Learning Objectives

By the end of this chapter, you will be able to:

• Describe and use the Snowflake architecture, including the Cloud Services layer, Compute layer, Database Storage layer, and the differences between Snowflake editions.
• Use Snowflake Interfaces and tools including Snowsight, the Snowflake CLI, and IDE integrations such as Visual Studio Code.
• Differentiate Snowflake object hierarchy and types across organization, account, and database scopes, including stages, schemas, tables, views, UDFs, file formats, stored procedures, pipes, shares, sequences, ML models, applications, and the rules of parameter precedence.
• Configure virtual warehouses across Snowpark Optimized and Standard generations, select the right size and scaling policy for each workload, and apply auto-suspend and concurrency guidance.
• Explain Snowflake storage concepts including micro-partitions, data clustering, the table types (Permanent, Temporary, Transient, Apache Iceberg, External, Dynamic), and view types (Standard, Materialized, Secure).
• Explain AI/ML and application development features including Snowflake Notebooks, Streamlit in Snowflake, Snowpark, Snowflake Cortex AI SQL functions, Cortex Search, Cortex Analyst, and Snowflake ML.

Executive Summary

• Snowflake separates Cloud Services, Compute, and Database Storage so that metadata management, query execution, and persistent storage scale independently, with each layer billed and tuned separately.
• The object hierarchy descends from Organization to Account to Database to Schema to schema-scoped objects (tables, views, stages, pipes, sequences, UDFs, procedures, ML models, applications), and parameters resolve from Account down to Session to Object scope.
• Virtual warehouses come in Standard (Gen 1 and Gen 2) and Snowpark Optimized variants, sized T-shirt style (XS through 6XL), with Standard or Economy scaling policies governing multi-cluster behavior.
• Micro-partitions are the immutable 50-500 MB compressed storage units that hold table data, and clustering metadata enables pruning so queries scan only the partitions that can satisfy the predicate.
• Snowflake Cortex, Snowflake Notebooks, Streamlit in Snowflake, Snowpark, and Snowflake ML are first-class surfaces inside the platform, executing against the same warehouses, storage, and governance model as SQL workloads.

Assumptions

• The reader has at least six months of hands-on Snowflake experience and general familiarity with relational database concepts, SQL, and a command line.
• All capacity, edition, and scaling values come from the most recent stable Snowflake documentation; verify edition-specific limits against the official documentation for your account's region and cloud provider.
• Examples use the fictional ACME_ANALYTICS account, SALES_DB database, and RAW, CURATED, and ANALYTICS schemas, with warehouses named like WH_ETL_M and WH_BI_S.