Which of the following solution options BEST addresses the business need in the most cost-effective manner?
Ensure that the Amazon Redshift cluster has been set up to make use of Auto Scaling groups with the nodes in the cluster spread across multiple Availability Zones.
Ensure that the Amazon Redshift cluster creation has been templated using AWS CloudFormation so it can easily be launched in another Availability Zone and data populated from the automated Redshift back-ups stored in Amazon S3.
Use Amazon Kinesis Data Firehose to collect the data ahead of ingestion into Amazon Redshift and create clusters using AWS CloudFormation in another region and stream the data to both clusters.
Create two identical Amazon Redshift clusters in different regions (one as the primary, one as the secondary). Use Amazon S3 cross-region replication from the primary to secondary region, which triggers an AWS Lambda function to populate the cluster in the secondary region.
Explanations:
While using Auto Scaling groups can help manage capacity, it does not inherently increase the reliability and availability of a Redshift cluster across Availability Zones, as Redshift does not support Auto Scaling in the same way as EC2. Also, data durability and recovery measures are not addressed by this option.
This option focuses on creating a backup and recovery strategy. By templating the cluster creation with AWS CloudFormation and using automated backups stored in Amazon S3, the cluster can be quickly restored in a different Availability Zone, meeting the requirement to restore within four hours.
This option involves streaming data to multiple clusters and does not directly address the reliability and availability of a single Redshift cluster. Additionally, creating a separate cluster in another region incurs higher costs and complexity without ensuring recovery within the four-hour requirement for a primary cluster failure.
Creating two identical Redshift clusters in different regions may provide redundancy, but it introduces significant cost and complexity. The requirement is to restore the cluster within four hours, which may not be feasible due to the replication and Lambda trigger setup. Additionally, cross-region replication could lead to data consistency issues.