Comment by jack9

Cut out Kafka by writing directly to S3 and bulk loading from S3 directory (optimal for Redshift). The article never details what "near real-time" means, which is bothersome.