z/VM CPU Pooling – A New Way To Help Manage Costs and CPU Usage for Linux Guests

December 22nd, 2015
Andy Hartman
Sr. Systems Engineer

 

IBM has added new functionality to z/VM 6.3 to help manage costs and processor resources used by Linux guests. This new functionality is called CPU Pooling and was made available in June of 2014 as a PTF. CPU Pooling allows you to group Linux guests together and set limits on how much processor capacity can be consumed by this group. CPU Pooling coexists with the setting of share values for individual guests, and the more restrictive setting takes president.

 

As the popularity of Linux has grown on IBM z Systems, over the past several years, the number of individual Linux servers (guests) has grown in both size and number running under z/VM. It can be challenging for some organizations to manage the various processor requirements placed on z/VM for certain types of situations and workloads. Having the ability to manage Linux guests in groups, with the ability to add and remove processor resources or guests from these groups, dynamically helps to manage the overall environment and meet service level agreements with various users and departments that share the z/VM environment.

 

CPU Pooling allows you to set limits on a group of Linux guests in one of two ways. You can set the amount of processor capacity as a percentage of all online processors available to the z/VM system, using the LIMITHARD operand of the DEFINE or SET CPUPOOL commands. You can also set a limit by the number of processors allowed to be used by this group using the CAPACITY operand on the DEFINE and SET CPUPOOL commands. z/VM assumes a 100% capacity on each processor available to the z/VM system, so from an LPAR level setting capping and weights on these resources will have affects on how the processor shares are determined, and they may differ from what is expected.

 

Using shares to affect processor usage, and other resources on z/VM, has been around for a long time and can be used to manage processor usage effectively, but it can become complex if your workload requirements change frequently. The other issue is managing workloads that consume excessive processor resources because of a problem with the software or middleware causing the Linux guest or guests to loop tightly affecting other Linux workloads. CPU Pooling allows you to more effectively manage these guests as a group… for instance, putting test or development workloads into a group that has limited processor resources limiting the impact of excessive processor usage on production guests.

 

Besides helping to manage processor usage, CPU Pooling also allows you to more effectively manage costs when it is used in combination with the IBM License Metric Tool. You can use sub capacity pricing for some IBM middleware running on Linux on z Systems. This could help to reduce costs by allowing you to only pay for the capacity that you are using for this middleware, as opposed to having to pay for the total capacity of all processors available to that software in a particular z/VM LPAR.

 

Mainline can help you plan for and implement CPU Pooling in your existing z/VM environment, and can help you understand it’s benefits, as well as the many other benefits of running Linux under z/VM on z Systems. Please contact your Mainline Account Executive to answer questions, or to get more information about CPU Pooling or using the IBM Licensing Metric Tool for sub capacity pricing.

Mainline