ON THIS PAGE
Virtual Network Interfaces for vMX
In a virtual environment, packet input and output capabilities play a significant role in the performance of the packet processing functionality inside the virtual machine, specifically the VFP VM. VFP supports two types of virtual network interfaces:
Paravirtualized—Paravirtualized network interfaces use network drivers in the guest OS and host OS that interact with the virtual environment and communicate effectively to give higher performance than fully emulated interfaces. In KVM, the supported paravirtualized interface is virtio. For VMware, VMXNET3 is supported.
PCI pass-through—PCI pass-through enables PCI devices such as network interfaces to appear as if they were physically attached to the guest operating system, bypassing the hypervisor and providing a high rate of data transfer. The physical network interfaces support single root I/O virtualization (SR-IOV) capability and can be connected to the VMs using PCI pass-through.
Choose the type based on how you want to use the vMX router. SeeTable 1.
Consideration |
Paravirtualization Technique |
PCI Passthrough Technique |
|---|---|---|
Interfaces |
virtio (for KVM), VMXNET3 (for VMware) |
SR-IOV |
Use Cases |
|
|
Host Requirements |
No requirements specific to this technique |
Physical NIC must support PCI passthrough |
VM Mobility (Junos OS Release 15.1F4 or earlier releases) |
Moving vMX instance to a new server without reconfiguration. |
Creating an identical vMX instance on a new server. |
Paravirtualization
Supported in Junos OS Release 15.1F4, in a paravirtualized router, the VM and the host work together to efficiently move packets from the physical NIC to the application in the VM. You implement paravirtualization on the vMX router by configuring virtio, a technique that the KVM hypervisor supports that optimizes network and disk operations for the VM. Both the VFP VM and the host contain virtio drivers that interact to move packets. You implement paravirtualization on the VMware server by configuring VMXNET3 on the ESXi hypervisor. You must provide the following information in the configuration file for each vMX interface:
Junos OS name
Unique MAC address
If you want to move the VM from one server to another, you can do so without reconfiguration, provided the names and MAC addresses of each interface remain the same.
PCI Passthrough with SR-IOV
Supported in Junos OS Release 15.1F4, The vMX router supports PCI passthrough in combination with single root I/O virtualization (SR-IOV). In the PCI passthrough technique, you directly assign a NIC’s memory space to a VM, enabling packets to bypass the hypervisor. Bypassing the hypervisor increases efficiency and results in high throughput of packets.
With SR-IOV, the hypervisor detects the physical NICs (known as a physical functions) and creates multiple virtual NICs (known as virtual functions) in the VFP VM. In the vMX implementation, the host dedicates a NIC to a single VM.
When you configure PCI passthrough with SR-IOV, you specify the following parameters for each vMX interface:
Junos OS name
Unique MAC address
Name of the physical NIC
Because you create a direct connection between a virtual NIC and a physical NIC, you cannot move a VM from one host to another. If you need to move a VM to another host, you must install a new vMX instance on that host, and delete the vMX instance on the original host.