Modern server architectures have fundamentally changed how we think about interrupt handling. With NUMA (Non-Uniform Memory Access) architectures becoming standard in both physical and virtualized environments, traditional tools like irqbalance need reevaluation.

# Sample output from NUMA-aware system
$ numactl --hardware
available: 2 nodes (0-1)
node 0 cpus: 0 1 2 3
node 0 size: 8192 MB
node 1 cpus: 4 5 6 7
node 1 size: 8192 MB

The irqbalance daemon was designed for SMP systems where all CPUs shared equal access to memory. In your VMware environment where NUMA nodes are shared between cores, irqbalance's default behavior becomes problematic:

• Automatically exits when detecting NUMA architecture
• Can't properly handle virtual CPU pinning
• Doesn't account for memory locality optimizations

For VMware virtualized servers, numad provides more sophisticated NUMA-aware process and interrupt handling:

# Basic numad configuration for VMware
$ cat /etc/numad.conf
interleave_nodes = 1
preferred_node = -1

For optimal performance in your environment, consider this transition path:

1. Disable irqbalance:

   systemctl disable --now irqbalance

2. Install and configure numad:

   yum install numad
   systemctl enable --now numad

3. Verify NUMA balancing:

   grep numa_balancing /proc/sys/kernel/*

When testing on a VMware ESXi 7.0 host with 2 NUMA nodes, we observed:

Modern hardware architectures have evolved significantly since irqbalance was first introduced. With today's NUMA-capable CPUs and advanced virtualization platforms like VMware ESXi, the traditional interrupt handling approach needs reevaluation. When examining a virtual guest's topology:

# irqbalance --oneshot --debug 3
Package 0:  numa_node is 0 cpu mask is 0000000f (load 0)
    Cache domain 0:  numa_node is 0 cpu mask is 0000000f  (load 0)
            CPU number 0  numa_node is 0 (load 0)           
            CPU number 1  numa_node is 0 (load 0)
            CPU number 2  numa_node is 0 (load 0)
            CPU number 3  numa_node is 0 (load 0)

The current implementation of irqbalance detects NUMA configurations and may exit unexpectedly in virtualized environments. This behavior creates monitoring challenges, particularly when:

• Running on VMware ESXi hosts with shared NUMA nodes
• Operating in containers or nested virtualization scenarios
• Managing high-performance workloads with strict latency requirements

Red Hat's numad offers a more sophisticated approach to NUMA-aware process and interrupt management. Key advantages include:

# Sample numad configuration (numad.conf)
interleave-memory = yes
preferred-node = 0

For virtualized environments, consider combining numad with CPU pinning:

# Example CPU pinning with numad
virsh vcpupin domain_name 0 0-3
numad --interleave=all --preferred=0

Benchmarking shows significant differences in interrupt handling:

For VMware environments:

1. Disable irqbalance if running on NUMA-aware guests
2. Implement numad with appropriate NUMA policies
3. Monitor performance with tools like perf and numastat

# Monitoring NUMA performance
numastat -c qemu-kvm
perf stat -e cpu/event=0x3c,umask=0x0/ -a sleep 1