With superior performance and the most hands-off model of ownership, Snowflake is the epitome of a data warehouse as a service. The model, cost, features, and scalability have already caused some to postpone Hadoop adoption. In its multicluster, scale-out approach, Snowflake separates compute from storage. It is fundamental to the architecture where multiple, independent compute clusters can access a shared pool of data without resource contention. Customers pay for what is used without a stairstep approach to resources and pricing.
- The cost model is simple at terabytes per year or computing hours.
- For primary storage, Snowflake uses Amazon’s Simple Storage Service (S3). Snowflake also uses an SSD layer for caching and temp space.
- Snowflake customers deploy a wide array of modern BI and visualization tools, some utilizing the ODBC and JDBC connectivity.
- Snowflake also offers a web interface.
- Snowflake SQL includes support of objects in JSON, XML, Avro, and Parquet using a special data type that can handle flexible-schema, nested, hierarchical data in table form.
- There are no indexes either, as zone maps are used for an abstract understanding of data in the database. SQL extensions include UNDROP and CLONE. Features include result set persistence and automatic encryption.
- No downtime is required for anything including upgrades or cluster expansion.
- Concurrency, a clear challenge in database scale-out, is a focus at Snowflake. Their automatic concurrency scaling is a single logical virtual warehouse composed of multiple compute clusters split across availability zones.
- Finally, Snowflake has a native connector for Spark built on the Spark Data Sources API.
Snowflake has jumped constraints found in databases from earlier development and honed a very promising cloud analytic database. Eminently elastic on a foundation of separation of computing and storage, Snowflake offers as close to a hands-off approach as we found. Snowflake is market-leading in what you would want for a multi-purpose cloud data warehouse analytical database.
Source – GigaOm Sector Roadmap: Cloud Analytic Databases 2017