Elastic computing refers to the ability to maximize performance by selectively activating and deactivating processor cores, which causes a subsequent increase or decrease in maximum frequency of the remaining active cores. This allows you to set the optimal balance between core count and maximum frequency for a given workload.
Each core supports one or two virtual processors, depending on whether hyperthreading is enabled or not. Virtual processors support threads; increasing the number of active cores increases the number of virtual processors, which allows the application to support more threads.
Deactivating all of the virtual processors associated with a core deactivates the core.
Some workloads are not able to take advantage of having many threads, but instead benefit from having fewer threads running at higher frequencies.
Before elastic computing, you had to order your server with a processor designed to have the optimal core count and frequency for your specific workload. Elastic computing gives you the ability to configure a single server dynamically to meet the needs of many types of workloads.
You can activate or deactivate cores using the Solaris or Oracle Linux OS, or the BIOS Setup Utility.
Using the OS commands, you can activate and deactivate virtual processors while the application is running, causing the corresponding cores to become inactive. This is the preferred method; however it requires that you have the Oracle Linux or Oracle Solaris operating systems. Using this method you can adjust performance dynamically, as often as necessary.
If you have any other supported operating system, you can activate or deactivate cores using the BIOS Setup Utility. While you can change the settings whenever you wish, this method requires rebooting the server.
The following table shows the relationship between active cores and maximum frequency.
|