CPU Thread Mapping Ratio

Disclaimer

This is not an official migVisor documentation and is provided for convenience purposes only. A general approach is described, and it may vary depending on specific use cases.

Applicable for Oracle and SQL Server only.

For database management systems running on a non x86 environment and with an x86 target environment, an additional field CPU Thread Mapping Ratio needs to be set when calculating the vCPU value for the database in Source Management.

The CPU Thread Mapping Ratio is the number of source system threads that are equivalent to one x86
thread. Industry benchmarks, such as SPECINT, are used to compare thread-to-thread performance for different chipsets to derive the ratio to use.

To find the CPU Thread Mapping Ratio for the server that the source database is running on and an instance type that can be used on Amazon RDS or EC2:

1. Find the server and non x86 processor models that host the database:

sudo dmidecode -s system-product-name

systeminfo | find "System Model"

lscpu

pxe/wmic CPU get NAME

2. Find the processor model of an x86 target environment in Amazon RDS or Amazon EC2 instance type.

3. Consult SPECINT or other benchmarks for the latest results and locate the processor models used for the database source and target.

If the specific models cannot be found, similar models can be used at the user’s discretion.

4. Use the Results/Base, Processor/Enabled Cores, and Processor/Threads/Core values for each processor type to calculate the Per Thread value for the source host and target instance:

   

Example

For source:
Per Thread = 118 / (8 * 2) = 7.375.

For target (simar to 2.4 GHz Intel Xeon® E5-2676 v3 (Haswell):
Per Thread = 601 / (20 * 2) = 15.025.

5. Find the CPU Thread Mapping Ratio:

   

The calculated CPU Thread Mapping Ratio can now be used for vCPU calculation under the Source Management → Key Sizing Metrics section.