Analysts use Redshift for recent curated data but occasionally join it with years of historical Parquet in S3. They want to query supported external tables without loading all historical files into local warehouse storage. Which capability fits?

1. COPY with the requirement that every historical row be loaded into local tables before any query.
2. Amazon Redshift Spectrum with an appropriate external schema and table metadata.
3. Create a normal view over a local table that does not contain the historical rows and expect it to discover the S3 files.
4. Use UNLOAD to move the recent local data to S3 and assume it creates the historical external query definitions automatically.