A 100-node HPC cluster in a VPC performs many DNS queries for RDS, S3, and on-prem resources accessible via Direct Connect. The cluster can scale 5–7x during peak events. The company currently uses two EC2 instances as the VPC DNS servers that forward queries to the VPC resolver and to on-prem DNS. DNS queries from nodes to resolve RDS and S3 endpoints are failing. What architectural change provides the most scalable DNS solution?

The correct option leverages AWS Route 53 Resolver, a highly scalable and managed DNS service, to handle the fluctuating demands of the HPC cluster. By configuring the HPC nodes to use the default VPC resolver (169.254.169.253), DNS queries for AWS services like RDS and S3 are resolved directly and efficiently within AWS, eliminating the current failures. Route 53 Resolver outbound endpoints and rules ensure that queries for on-premises resources are correctly forwarded to the on-prem DNS servers via Direct Connect. This fully managed approach removes the operational overhead and scaling limitations of EC2-based DNS servers. Scaling out or up EC2 instances (options 1 and 2) introduces manual management, potential bottlenecks, and does not address the fundamental issue of efficiently resolving AWS service endpoints. Creating Route 53 Resolver inbound endpoints (option 4) is for allowing on-premises resources to resolve DNS names in AWS, not for AWS resources to resolve on-premises names, and would still require HPC nodes to query on-prem DNS for all lookups, which is inefficient for AWS services.