Does NUMA architecture cause a problem when running Linux virtual machines (VMs) on Hyper-V?
It can. Non-uniform memory access, or NUMA, is a computer design approach that basically segregates memory into areas or spaces called NUMA nodes. The idea is that NUMA architecture can boost performance by allowing a processor to access its own local NUMA node -- which is really memory space -- faster than it can access memory in other NUMA nodes that other processors in the system use. Almost all modern servers employ NUMA architecture in the processor/memory relationship.
NUMA support is normally enabled by default, but Linux VMs may experience problems with NUMA support in several conditions. Performance problems may occur if the Linux VM uses more than seven virtual processors or more than 30 GB of memory. These are recognized issues with the 2.6.x Linux kernels.
If either condition exists on your system, add the numa=off line to the GRUB boot.cfg file and reboot the VM. This should disable the use of NUMA for the VM and cause the VM to treat all memory as a single uniform entity. While this causes a slight performance penalty, it should alleviate any broader stability or reliability issue.
Hyper-V does support Linux VMs across multiple distributions, but feature support and stability can be far from uniform -- and support for generation 2 VM features may be impossible under almost all Linux versions. Adopting Linux VMs should include ample proof-of-principle testing and performance benchmarking in order to validate the Linux VM before rolling it out to a production data center. Similarly, any changes to the Linux distribution or underlying Windows integration services should be preceded by a generous testing phase to identify and resolve potential problems.
Problems to prep for when running Linux VMs in Hyper-V
Planning the live backup of an Oracle Linux VM in Hyper-V
Why SUSE VMs have UEFI and secure boot issues
Dig Deeper on Microsoft Hyper-V management
Related Q&A from Stephen J. Bigelow
Get to know VMware vSphere's Admission Control tool and use it to reserve the resources necessary for VM failover with cluster resource calculations ... Continue Reading
Use heartbeats, VM monitoring and application monitoring to fully examine the causes of VM unresponsiveness. Adjust sensitivity levels to focus on ... Continue Reading
Combine Distributed Resource Scheduler and vSphere High Availability to design balanced failover clusters. Pay attention to affinity rules, which can... Continue Reading
Have a question for an expert?
Please add a title for your question
Get answers from a TechTarget expert on whatever's puzzling you.