GPU & SR-IOV

CobaltCore supports passthrough of GPUs and SR-IOV network interfaces to virtual machines for workloads that require direct hardware access - machine learning inference, HPC, and latency-sensitive networking.

Documentation in progress

This section is being expanded. The information below describes the intended architecture and current capabilities.

GPU passthrough

GPU passthrough uses VFIO (Virtual Function I/O) to assign a physical GPU directly to a VM, bypassing the hypervisor's virtualization layer. The VM sees the GPU as if it were physically attached.

Requirements:

• CPU and motherboard with IOMMU support (Intel VT-d or AMD-Vi) enabled in BIOS
• VFIO kernel modules loaded on the GardenLinux hypervisor
• GPU bound to the vfio-pci driver rather than the native driver

OpenStack integration: Nova exposes GPU resources as traits (CUSTOM_GPU_*) through the placement API. Cortex placement policies can target GPU-capable hypervisors for VM scheduling.

SR-IOV networking

SR-IOV (Single Root I/O Virtualization) allows a single physical NIC to present multiple virtual functions (VFs) to VMs, each with near-native throughput and latency.

Requirements:

• SR-IOV capable NIC with SR-IOV enabled in BIOS
• sriov-device-plugin deployed on hypervisor nodes
• Neutron SR-IOV agent configured

OpenStack integration: Neutron manages SR-IOV ports. VMs with SR-IOV ports bypass OVN and receive a VF directly.

See also

• Compute - Hypervisor
• Networking - OVN