check-circle-line exclamation-circle-line close-line

Oracle database workloads on VMware Cloud on AWS provide high performance, achieving hundreds of thousands of orders per minute. The scale-up and scale-out performance shows that Oracle running in this environment is capable of supporting big workloads as well as a large number of small- to medium-sized workloads as well.  Most of the best practices that have been recommended for Oracle on vSphere still apply to this new environment. The VMware-published Oracle Databases on VMware Best Practices Guide is the best source for complete details on best practices.

Distributed Resource Scheduling

Distributed Resource Scheduling (DRS) is a feature of vSphere that decides where to place VMs upon power-on, and when to move running VMs. This is the VM load balancing feature that is used to make sure that VMs are placed so that performance is optimal across a vSphere cluster. When using VMware Cloud on AWS, this feature is turned on and configured as part of the service. VMs are placed and moved around the cloud automatically.

Initial placement of the VMs was fairly easy for DRS in the test cases where there were enough resources—this was all of the test cases with 8 VMs or less. In cases where there were more, DRS would sometimes decide to move VMs around after load was applied to the VMs, meaning that, in some of the tests with more than 8 VMs, multiple vMotion migrations occurred during the first test run with that number of VMs. This sometimes caused the initial set of tests to have lower than expected throughput in the form of low OPMs. Once the vMotion migrations completed, throughput increased back to expected levels. 

VM Sizing

It is important to size the VM with the amount of virtual CPUs and memory that are needed for that workload.  Creating VMs that are larger than they need to be can result in wasted resources and lower overall performance across all VMs. Looking at the results of the 16 vCPU VMs compared with the 18 vCPU VMs, it is clear that better overall performance was possible with more slightly smaller VMs. At the same time, an 18 vCPU VM did outperform a 16 vCPU VM. Rightsizing the VMs so that they have the resources they need can lead to overall better utilization.


The performance scalability that we observed in these tests on the VMware Cloud on AWS environment were very similar to what is seen with Oracle database test workloads running on VMware vSphere onsite. Because the software stack is essentially the same and the key difference is location and deployment methodology, this was expected. High-performance Oracle databases can be run successfully with good performance on VMware Cloud on AWS.