Intermittent VM performance differences in a cluster

Products

• VMware vSphere ESXi

Issue/Introduction

• Intermittent performance differences across virtual machines (VMs) in a particular cluster.
• These symptoms often manifest as elevated CPU Ready (%RDY), Co-Stop, or CPU Wait Time values.
• Distributed Resource Scheduler (DRS) is set to Fully Automated mode for the cluster.

Environment

VMware vSphere ESXi 7.X

VMware vSphere ESXi 8.X

 

Cause

The root cause of these random VM performance differences is typically a combination of the following:

• Hyperthreading disabled on some hosts: Even though a host that is enabled for hyperthreading ought to behave similarly to a host without hyperthreading, it does allow a single processor core to execute two independent threads simultaneously. This guarantees that load is spread smoothly across processor cores in the system. While hyperthreading does not double the performance of a system, it can definitely increase performance by better utilizing idle resources leading to greater throughput for certain important workload types.

• DRS in Fully Automated mode: DRS may migrate workloads to hosts with lower logical CPU capacity (i.e., those with Hyperthreading disabled), inadvertently increasing CPU wait times that may result in slight performance differences on VM's.

Resolution

To resolve this issue and ensure consistent performance across the cluster, enable Hyperthreading on all ESXi Hosts:

• Reboot each affected host and enter the BIOS/UEFI setup.
• Locate the Hyperthreading setting and enable it.
• Save changes and reboot the host.
• Verify Hyperthreading status in vSphere by navigating to: Host > Configure > Hardware > Processor > Hyperthreading

Link for reference: Enable Hyperthreading