11
If query patterns shift toward frequent joins with core Redshift tables, revisit that decision. Spectrum queries are noticeably slower for anything beyond simple scans and filters, mostly due to S3 read latency.
New dataset is queried maybe a few times a week, moderate size around 500GB. Debating whether to load it fully into Redshift or query it in place from S3 via Spectrum.
If query patterns shift toward frequent joins with core Redshift tables, revisit that decision. Spectrum queries are noticeably slower for anything beyond simple scans and filters, mostly due to S3 read latency.
Spectrum is a good fit for exactly this profile. Infrequently queried data doesn't justify the storage and load cost of bringing it fully into cluster storage.
Sign in to reply.