Migrate SQL Server to AWS RDS or EC2

One of the world’s largest multimedia content providers was looking for a way to increase the level of user satisfaction by improving the performance of their enormous database system. They wanted to leverage operational best practices to increase resiliency/uptime, performance, and scalability at the lowest possible cost. Their overall goal is to move all of their applications out of their datacenter and over to Amazon Web Services through a lift-and-shift process.

The first consideration we needed to help our customer through was determining which workloads should move to SQL Server running on EC2 instances versus SQL Server on RDS. AWS does provide a white paper Deploying Microsoft SQL Server on Amazon Web Services that features SQL Server 2014 written in 2016. Since then, Microsoft has shipped two new versions of SQL Server and AWS has made 12 major improvements to RDS for SQL Server including support for SQL Server 2017.

Considerations for RDS

Based on the criteria we gathered for high availability, disaster recovery, performance, scale, and SQL Server features used within the application, our default is to always consider RDS as the deployment target because of the great platform as a service features it provides. We then take a look at the following considerations:

• Are there any features that applications use that are not supported by AWS RDS for SQL Server? We use the reporting capabilities for the AWS Schema Conversion Tool (SCT) to identify incompatible features. AWS does a great job of updating to align with the ever-shrinking list of unsupported SQL Server features with RDS.
• Can we reduce the administrative burden with features like automatic backups, automatic patching, seamless replication, and super-easy point-in-time restore?

Considerations for SQL Server on EC2

• Use of unsupported features. SCT provides a report of the unsupported features and offers suggestions on how to work around the issues. If the workaround has little impact on the application code, then we will consider it based on the cost to deliver the change. On the other hand, if the workaround is significant, going with EC2 may make sense.
• Hardware resources are not sufficient to support performance and high availability. RDS instances are quickly aligning with the largest of instance types. As of July 2018, SQL Server 2016 & 2017 RDS on Enterprise Edition supports the db.r4.xlarge–16xlarge instance type with 64 vCPUs, with 195 ECUs, and 488 GiB. However, if a larger machine is required or there is a higher level of availability that requires a feature like Always On Cluster Failover Instances and/or Availability Groups, then EC2 is the answer.