Data center overlay

Currently, the main overlay technologies include VXLAN, NVGRE, STT, and spb mac-in-MAC, TRILL, Cisco OTV and LISP. These technologies basically use encapsulation or tunneling to create virtual networks.
VXLAN uses MAC-in-UDP encapsulation to overlay a layer-2 network to implement virtual network. Currently, VXLAN is a subset of the IETF standard draft, supporting vendors such as VMware, Cisco, Arista network, brocade, RedHat, and Citrix. IBM also supports VXLAN. Specifically, VXLAN uses MAC-in-UDP encapsulation, including 24-bit network identifiers, 50 bytes overhead, and can provide 0.16 million logical network addresses. For VMS, VXLAN is transparent and can still communicate with MAC. Theoretically, all endpoints in the same VXLAN network segment can communicate with each other, even if the two endpoints belong to different IP subnets. This solves a problem. VM migration and use in a large network without using the layer-2 construction technology.
NVGRE is very similar to VXLAN and uses encapsulation to carry L2 Networks on an existing L3 IP network. Only NVGRE adopts the GRE encapsulation instead of UDP encapsulation. In the opinion of Microsoft, the major support vendor, some network performance problems can be solved through the intelligent network card, that is, the NVGRE endpoint processing on the Smart Nic to uninstall the hypervisor vSwitch. The smart NIC also has APIs integrated with the overwrite controller and hypervisor management system. There is a controller that plays an important role in solving some problems encountered in the control plane.
STT is a technique proposed by Nicira. The STT protocol uses stateless TCP headers in the IP header, allowing the end system's tunnel endpoint to use the TCP multipart unload function (TSO) of the TCP unload engine (TOE) residing on the server Nic ).
SPBM uses IEEE 802.1ah MAC-in-MAC encapsulation and IS-IS routing protocol to provide layer-2 network virtualization through VLAN extension, in addition, there is usually a non-loop equivalent cost multi-path two-layer forwarding function related to SPB.
VLAN extensions can be implemented by a 24-bit virtual service network (VSN) instance service ID (I-SID), which is part of the outer MAC encapsulation. Unlike other network virtualization solutions, there is no need to change either the hypervisor vSwitch or Nic, or the existing switch hardware that supports IEEE 802.1ah MAC-in-MAC encapsulation. For SPBM, the control plane IS provided by the IS-IS routing protocol. As described in the draft IP/spb ietf, SPBM extensions can be added to support layer-3 forwarding and layer-3 virtualization by using the IP encapsulation of the outer spbm mac. This draft defines how SPBM nodes execute Inter-ISID or Inter-VLAN routing. In addition, IP/SPB can provide layer-3 vsns across SPBM networks by extending virtual routing and virtual Forwarding (VRF) instances at the edge of the network, without supporting core Switches of VRF instances. The VLAN-extended vsn and the vrf-extended VSN can run on the same SPB network at the same time, providing isolated layer 2 and Layer 3 traffic for multi-tenant environments. With SPBM, all core switches that begin to define the SPBM boundary on the access or aggregation switch must be enabled by SPBM. Currently, available SPBM hardware switches include Avaya and Alcatel-Lucent.
TRILL (Transparent Interconnection of Lots of Links) is a protocol that applies the layer-3 link status routing technology to L2 Networks. TRILL implements L2 routing by extending the IS-IS routing protocol, which can meet the big Layer 2 networking requirements of data centers and provide solutions for data center businesses.
OTV optimizes the WAN or Internet expansion of VLANs in the data center and uses MAC-in-IP encapsulation. It uses the IS-IS routing protocol extension to advertise the accessibility of MAC addresses to prevent the unknown flood of traffic from the destination endpoints on the WAN.
LISP is an IP-in-IP Encapsulation technology that allows the system to retain IP addresses (IDS) even when the end system is migrated to different subnets ). With the migration of lisp vms, IP endpoints, such as VMS, can be migrated to any place without having to worry about what their IP addresses are, while maintaining direct route of client traffic. Using ving VRF to the LISP instance ID to create a layer-3 Virtual Network, LISP can also support multi-tenant environments.