Overview of Amazon Redshift Spectrum. How does Redshift Spectrum work? It almost seems like magic. Essentially, the data stored in S3 is formatted like a Redshift table and cataloged with something like AWS Glue. That's the high-level explanation of what happens. Amazon Redshift Spectrum nodes are dedicated Amazon Redshift Servers managed by AWS that are independent of customer-provisioned clusters. Because of this, Redshift Spectrum Queries use much less of a cluster's processing capacity than other queries because compute-intensive activity is pushed into the spectrum nodes.

Based on the demands of queries, Redshift Spectrum can potentially scale to use thousands of spectrum nodes to take advantage of Redshift's massively parallel processing architecture. Redshift Spectrum Tables are created by defining the structure of external files and then registering them as tables in an external data catalog. You can think of it as an external table. The data catalog can be AWS Glue. The data catalog that comes with Amazon Athena or an Apache Hive Metastore.

External tables can be created and managed, either by using Redshift's data definition language commands, or with any other tool that can connect to the external data catalog. Changes to the external data catalog are immediately available to Amazon Redshift. Optionally, the external tables can be partitioned on one or more columns. Though, it's a good idea to partition data. When Redshift executes a query, it creates a query plan.

If the data is partitioned, the query plan will know what data it needs and, just as importantly, know what data to skip. This means defining partitions as part of the external table can improve performance. After the Redshift tables have been defined, tables can be queried and joined, like any other Redshift table. Keep in mind that these tables are read only. Update operations are not possible. Redshift Spectrum Tables can be added to multiple Amazon Redshift clusters. This means the same data stored in S3 can be queried by any Redshift cluster in the same AWS region. When the Amazon S3 files are updated, the data is immediately available for query.