Before considering where to place SQL Server workloads on a cloud platform, ensure your VM container is rightsized. A workload’s performance profile should be collected over a sufficient period of time to reflect applications spikes in resource utilizations. While defining the required time range to collect time series data, consult with DBAs and application owners to understand the workload profile. At least a full month of “non-rolled up” time series data is recommended prior to execute the performance analysis.
Utilizing Blue Medora SQL Server management pack with vRealize® Operations Manager™ is proven to be very helpful in this preparation phase. While analyzing captured data, make sure your rightsizing approach has been agreed upon by administrators, applications owners and business owners, and that it comprehends both spikes (high performance) and average utilization (higher density).
The following should be considered while sizing SQL Server workloads:
For CPU and memory resources allocation, check the availible host configurations for VMware Cloud on AWS to verify the workload will fit and not overcommit host resources.
Account for differences in physical CPU architectures between your current environment and the host instances used in VMware Cloud on AWS.
Always size the CPU resource based on the actual workload, as vCPU can be easily added later.
The storage layer in VMware Cloud on AWS is provided by VMware vSAN - hyperconverged infrastructure (HCI) solution, if using an Amazon EC2 I3.metal instance. Adding storage will require the addition of compute resources (hosts) as well. As an alternative, for workloads with the primary capacity requirements use Amazon Elastic Block Store (EBS) with Amazon EC2 R5.metal hosts. An I3 instance still should be your primary choise for a performance OLTP workloads.