Complete Guide on network Virtualization (1)

Many new and recommended protocols have emerged for how to optimize the data center Ethernet and support its provision of server virtualization. Some of the protocols aim to achieve network virtualization by creating multiple virtual Ethernet networks that can share the same physical infrastructure. The sharing method is similar to that of multiple virtual machines sharing the same physical server.

Most protocols applicable to network virtualization basically use encapsulation and tunneling technology to create virtual network coverage. The most discussed protocols in the industry include VXLAN, NVGRE, STT, and spb mac-in-MAC. SPB is already an IEEE standard, and VXLAN is the most likely standard among various protocols that may become an IETF standard.

Traditional Network Virtualization

One-to-multiple Virtualization of network entities is not a new concept. The most common examples are VLAN and VRF (Virtual routing and forwarding ).

A VLAN can divide the network into a maximum of 4094 broadcast domains, and specify a 12-bit vlan id for each broadcast domain in the Ethernet header. VLAN is a convenient way to isolate different types of traffic in a shared LAN infrastructure.

Data centers use server virtualization in large quantities, and restrictions on the number of VLANs may cause problems. Especially when a large number of tenants need support and each tenant needs multiple VLANs. With the 802.1Q trunk link, you can expand VLANs in the data center to support the mobility of VMS, but this increases operating costs and complexity. Even in a data center connected to a two-tier server-server, a large number of VMS each have their own MAC address, it will also burden the forwarding table function of the two-tier switch.

VRF is a type of layer-3 network virtualization. Physical routers support multiple vro instances. Each instance runs its own route protocol instance and maintains its own forwarding table.

Unlike VLAN, VRF does not use tags in the header to specify a specific VRF for each group. In each hop, appropriate VRF is obtained based on the input interface and frame information. Another requirement is that a VRF instance must be configured for each intermediate router in the end-to-end path of the data packet to forward the data packet.

Use Coverage Network Virtualization

Due to defects in traditional VLAN or VRF models, many new technologies have emerged to create virtual networks. Most of them use encapsulation and tunneling technology to build multiple virtual network topologies on the same physical network through coverage.

A virtual network can be a 2-layer or 3-layer network, while a physical network can be 2-layer or 3-layer, or a network that combines the two, this depends on the coverage technology used. With the coverage technology, the outer (encapsulation) header contains a 24-bit long domain, carrying a virtual network instance ID (VNID), and assigns a virtual network to the data packet to be forwarded.

The coverage of virtual networks provides many benefits, including:

● Supports virtual networks with basically no limit. For example, a 24-bit header can create up to 16 million virtual networks.

● Decoupling virtual network topologies, service categories (L2 or L3), and physical network addressing. This decoupling prevents problems such as the large MAC table on a physical switch.

● Migration of virtual machines is independent of physical networks. If a VM needs to change its location or even migrate to a new subnet, The vswitch that overwrites the edge only needs to update its ing table to reflect the new location of the VM. The network of the new VM can be pre-configured at the edge of the network.

● The ability to manage multiple tenants to overwrite IP addresses.

● Multi-path Forwarding is supported in the virtual network.

The main difference between various coverage protocols lies in the Encapsulation Format and the functionality of the control plane, that is, the entry (encapsulation) device is allowed to map a frame to an appropriate exit (disassembly) device.