VMware’s unsupported version of its ESXi hypervisor for Arm platforms has been updated with support for two-socket servers and Nvidia's Arm-flexing development boards.
The previous version of ESXi-on-Arm “Fling” that VMware offers to the curious willing to work without a safety net added support for Ampere’s server-class Arm processors. Now version 1.4 adds support for Ampere’s Mount Jade reference platform for two-socket servers. VMware hasn’t yet got the code running across NUMA nodes, so even though Mount Jade can handle a pair of Ampere’s 80-core Altra processors, it’s only possible to create a VM that uses half of the total possible core count “at this time.”
- VMware adds support for 80-core Ampere Altra chips to its experimental Arm hypervisor
- Nvidia exec love-bombs Arm's licensing model, almost protests too much
- India’s massive COVID-19 wave slows VMware desktop hypervisor development
- Two ransomware strains target VMware’s ESXI hypervisor through stolen vCenter creds
Nvidia admirers also have more reason to consider ESXi-on-Arm as the GPU pioneer’s Tegra Xavier AGX and Tegra Xavier NX boards have also gained support in the experimental hypervisor.
The AGX is designed for embedded and industrial application development while the NX aims to bring AI to edge devices. VMware stated, almost apologetically, that support for the devices’ GPUs is not enabled “at this time”.
Other new bits in ESXi-on-Arm include:
- Improved physical memory unit (PMU) virtualization
- Fixed virtual AHCI support for some ACPI OSes
- Improved time virtualization
- Improved VM performance for multi-socket Arm servers
- Fixed virtual NVMe support in UEFI and some OSes
- Improved interrupt controller virtualization
- Improved compatibility with newer guest OS Linux kernels
- Improved USB stability issues, especially with RTL8153-based USB NICs and especially on Raspberry Pi and Tegra Xavier.
VMware’s loaded up its announcement post for the new software with disclaimers that its current limitations aren’t permanent. Clearly, Virtzilla is considering applications for its hypervisor on Arm well beyond the discrete Project Monterey effort aimed at getting its hypervisor running on SmartNICs. ®