View Only

Understanding Integrated Facility for Linux (IFL)

By Pete Heyrman posted Tue June 16, 2020 02:42 PM

PowerVM IFL LogoOn enterprise class servers, IBM supports an option to purchase processors and memory for Linux workloads at lower costs than traditional workloads.  With the Integrated Facility for Linux (IFL) option, customers have the flexibility to run new Linux workloads on the same servers as their existing AIX and IBM i workloads.  IFLs can also be used as an option to create a Linux only enterprise class server for new applications like SAP HANA.  At a later point in time, if the need arises, additional cores capable of running AIX or IBM i could be licensed.

What is considered an IFL-compliant workload?
Prior to FW830 on POWER8 servers, IFL-compliant workloads were restricted to Linux only partitions (SUSE, RedHat and such). 

Starting with FW830 on POWER8 and all POWER9 enterprise class servers, IBM considers Linux and VIOS partitions as IFL-compliant.  This allows customer to utilize all the virtualization features of the VIOS and PowerVM for Linux workloads.

HMC information about IFL cores
On a server with IFL cores enabled, the HMC will display the following when the System – Processor, Memory, I/O panel is selected:

PowerVM IFL ExampleFor this server there are 4 cores limited to Linux or VIOS usage and 4 cores licensed for usage by any OS (AIX, IBM i, Linux or VIOS).  Since IFL-compliant workloads can be dispatched on all core types, IFL-compliant workloads have 8 total cores available.  Similar information can also be retrieved through the HMC command line interface using the command lshwres -m YOUR-SERVER-NAME -r proc --level sys.

Management of IFL cores
When a server has a mixture of IFL and non-IFL cores, AIX and IBM i partitions are not allowed to use IFL resources, but IFL-compliant workloads can run on any core.

Prior to FW840 on POWER8 servers, customers are responsible to ensure that resources assigned to workloads that were not IFL-compliant do not use any IFL resources.  One way for customers to manage this restriction would be to put their AIX and IBM i partitions in a shared processor pool to ensure these partition types are not using any IFL resources. 

Starting in FW840 on POWER8 and all POWER9 enterprise class servers, IBM has automated the tracking of cores used by AIX and IBM i.  For example, if you purchase a server with 20 cores for AIX/IBM i and 20 cores of Linux, the PowerVM hypervisor will ensure automatically that at most 20 AIX/IBM i virtual processors will be running at any point in time.  This removes the need for customers to use techniques like shared processor pools to manage CPU consumption.  Also, if a dynamic LPAR (DLPAR) operation is requested for an AIX/IBM i partition to add physical processors to a dedicated partition or entitled capacity to a shared partition, the request will be allowed if the server remains in compliance with the licensed cores for AIX/IBM i.  Similarly, if a partition attempts to boot and there are insufficient AIX/IBM i cores available, the partition may boot with reduced resources or may not boot at all depending on the minimum resources set for the partition.  This is all part of the automatic compliance that has been built into the PowerVM hypervisor.

Live Partition Mobility when target server has IFL cores
When migrating an AIX/IBM i partition to a server that contains IFL cores, the request will only succeed if there are enough available AIX/IBM i cores to satisfy the resources required by the migrating partition.  Since Linux partitions can run on any type of cores, migration only needs to ensure there are enough processors of any type available on the target server.

IFL resources and Affinity
The PowerVM hypervisor will automatically select resources that optimize the resource affinity for partitions independent of the difference between IFL and non-IFL resources.  For example, if you have an IFL-compliant partition with 0.5 processor units and an AIX partition with 0.5 processor units, the hypervisor can share the same physical core if this optimized the affinity for the partitions.  The hypervisor manages memory is a similar manner to optimize the partition affinity to deliver optimum performance.

This overview of IFL resources provided an insight into how to leverage your enterprise class server if you are considering using the IFL option to reduce your total cost of ownership.  You can read more about IFL here.

Contacting the PowerVM Team
Have questions for the PowerVM team or want to learn more?  Follow our discussion group on LinkedIn IBM PowerVM or IBM Community Discussions