Finetuning and Best Practices¶

vCPU and Scheduling¶

• Start with two vCPUs for small OpenVMS instances; only add more if you see sustained CPU demand.

• Avoid over-provisioning vCPUs. Too many vCPUs can increase scheduling latency and make interactive sessions feel slower.

• For latency-sensitive workloads (terminal responsiveness, DECnet routing, etc.), prefer fewer, busy vCPUs rather than many mostly idle ones.

Memory¶

• Allocate enough RAM to keep OpenVMS from heavy paging; monitor pagefile usage inside OpenVMS.

• Avoid aggressive memory overcommit on the ESXi host for OpenVMS production VMs.

• Keep memory reservations for critical OpenVMS systems if the host is heavily loaded, so they are not ballooned or swapped by ESXi.

Storage¶

• Use a paravirtual SCSI controller type that is supported starting with VSI OpenVMS x86-64 Version E9.2-4.

• Place system disks for OpenVMS on storage with low latency and good IOPS, such as NVMe SSD.

• PCI passthrough for Fibre Channel provides the fastest performance for shared cluster data disks.

Network¶

• Use a single vNIC for simple test systems; add separate vNICs for management, DECnet, and application traffic if you need isolation.

• Assign vNICs to separate physical NICs for higher performance.

• Ensure the vSwitch and port group settings match the VLAN and MTU of your physical network.

Console and Serial Behavior¶

Document the exact console connection method (pipe name, port, host), so operations staff can reach it during outages.

For more information on performance tuning, see VMware documentation.