Lead Image © Joe Belanger, 123RF.com
Best practices for KVM on NUMA servers
Tuneup
Non-uniform memory access (NUMA) [1] systems have existed for a long time. Makers of supercomputers could not increase the number of CPUs without creating a bottleneck on the bus connecting the processors to the memory (Figure 1). To solve this issue, they changed the traditional monolithic memory approach of symmetric multiprocessing (SMP) servers and spread the memory among the processors to create the NUMA architecture (Figure 2).
Figure 1: Traditional SMP architecture.
Figure 2: NUMA architecture.
The NUMA approach has both good and bad effects. A significant improvement is that it allows more processors with a corresponding increase of performance; when the number of CPUs doubles, performance is
...
Buy this article as PDF
(incl. VAT)
Buy ADMIN Magazine
US / Canada
UK / Australia
Related content
-
Getting your virtual machine dimensions right
Massive, performance-hungry VMs require proper handling to meet their dynamic requirements. We give you some rules to help size these monsters properly. -
Virsh Libvert Tool
With the command-line tool virsh, a part of the libvirt library, you can query virtual machines to discover their state of health, launch or shut down virtual machines, and perform other tasks – all of which can be conveniently scripted.
-
CPU affinity in OpenMP and MPI applications
Get better performance from your nodes by binding processes and associating memory to specific cores. -
Processor and Memory Affinity Tools
Get better performance from your nodes by binding processes and associating memory to specific cores .
-
Visualizing kernel scheduling
The Google SchedViz tool lets you visualize how the Linux kernel scheduler allocates jobs among cores and whether they are being usurped.