Question

i need help writing a outline for my research paper on the comparison and contrast of NVGRE,VLAN, and STT?

Answer 1

With the time, there are lot of new encapsulation frame formats purpose-built for the data center, including Virtual Extensible LAN (VXLAN), Network Virtualization Using Generic Routing Encapsulation (NVGRE), Transparent Interconnection of Lots of Links (TRILL), and Location/Identifier Separation Protocol (LISP). So as we go through this article you will came to know some of the overlay transport protocols and they are

VXLAN: Virtual Extensible LAN
OTV: Overlay Transport Virtualization
LISP: Locator/Identifier Separation Protocol
NVGRE: Network Virtualization Using Generic Routing Encapsulation
STT: Stateless Transport Tunneling
Apart from the underlay network(IP network), Network overlays are virtual networks of interconnected nodes that share an underlying physical network, allowing deployment of applications that require specific network topologies without the need to modify the underlying network.

With the help of Network overlay, you can optimise the device functions and also reduces the complexity of the network devices. In the case of server-based overlays, this function is implemented on the server. In the case of network-based overlays, this function is implemented on the first switch (at the top of the rack). With the help of Ovelay networks you can achieve and provide scalable Layer II networks for a multitenant cloud that extends beyond 4000 VLANs. This capability is very important for private and public cloud hosted environments.

You can also maintain the fabric scalability and flexibility, because the overlay virtual network no longer needs to be constrained to a single physical location. The overlay encapsulation also allows the underlying infrastructure address space to be administered separately from the tenant address space.

NVGRE-Network Virtualization Using Generic Routing Encapsulation

Well, NVGRE allows the creation of virtual Layer 2 topologies on top of a physical Layer 3 network. With the help of NVGRE you can achieve by tunneling Ethernet frames inside an IP packet over a physical network. NVGRE supports a 24-bit segment ID or virtual subnet identifier (VSID) similar to VXLAN, providing up to 16 million virtual segments that can uniquely identify a given segment.

The difference between VXLAN and NVGRE is that NVGRE header includes an optional flow ID field. In multi-pathing deployments, network routers and switches that can parse this header can use this field together with the VSID to add flow-based entropy, although this feature requires additional hardware capabilities.

Host VM VM VM Virtual Switch GRE PA GRE PA Virtual Switch VM VM Host VM Outer MAC (188) Original Ethernet Frame Outer IPv4 GRE with VSID

Fig 1.1- NVGRE Header format

NVGRE, header has outer MAC of 18 bytes; outer IP of 20 bytes when using IPv4 and GRE with VSID of 16 bytes.

As with VXLAN, the NVGRE draft standard does not specify a method for discovering endpoint reachability. Rather, it suggests that this information can be provisioned through a management plane or obtained through a combination of control-plane distribution or data-plane learning approaches.

Note: Microsoft is one of the major vendor who uses NVGRE concept and still a lot of vendors are doing their efforts to develop NVGRE like Cisco, VMware and Juniper.

STT Protocol- Stateless transport tunnelling

Well, now you understand basics of NVGRE protocol, similar to NVGRE we have another protocol named as STT and is an overlay encapsulation scheme over Layer 3 networks that use a TCP-like header within the IP header. The use of TCP fields has been proposed to provide backward compatibility with existing implementations of NICs to enable offload logic, and hence STT is specifically useful for deployments that are target end systems for example a virtual switches on physical servers.

Physical Host A Physical Host B VMs Initiator 1 Initiator 2 Initiator 3 Initiator 4 Initiator 5 Initiator 6 Initiator 7 Initiator 8 Initiator9 VMs Target 1 Target 2 Target 3 Target 4 Target 5 Target 6 MLAG 10G LAG L2 Switch A LAG 10G Tunnel L2 Switch B Bond Bond Target 8 Target 9

Fig 1.2- STT tunneling

STT is designed for a specific reason and is specifically addresses the issue of size mismatch between Ethernet frames and the maximum transmission unit (MTU) supported by the underlying physical network. Most of the end host operating systems set the value of the MTU at a small size so that the entire frame plus any additional (overlay) encapsulations can be transported over the physical network.

This setting may result in a potential performance degradation and additional overhead compared to frames that can be transmitted with their desired maximum segment size (MSS). STT seeks to exploit the TCP segmentation offload (TSO) capabilities built into many NICs today to allow frame fragmentation with appropriate TCP, IP and MAC address headers, and also the reassembly of these segments on the receive side.

Lets compare these protocols with VXLAN and LISP protocol to understand more where and what to use when we have the requirement of the overlay networks.

VXLAN Overlay Network Technology NVGRE Overlay Network Technology Overlay Network Technology Overlay Network Technology 24-bit virtual subnet identifier-VSID with optional STT LISP Features Technology Overlay Identification 24-bit virtual network ID (VNI) 8-bit flow ID 64-bit context ID 24-bit LISP instance ID UDP encapsulation, port GRE based, protocol type UDP based port 8472, Adds 8 Ox6558( transparent and TCP based and adds 8 byte 4341 and adds 8 byte LISP byte VXLAN header 50 bytes 50 bytes MTU should be increased 50 MTU discovery and setting bytes to avoid fragmentation DF bit on outer header to t uses interface MTU and Stateless and Stateful LISP of VXLAN packets Encapsulation Overhead Network MTU ethernet bridging) 42 bytes 42 bytes STT header 76 bytes 64KB header 50 bytes 50 bytes avoid fragmentation No Microsoft Hyper-V virtual Switch not decided yet methods No Cisco Nexus 1000v Virtual Switch not decided yet Fragmentation Offload to NIIC No TCP segmentation Yes Cisco Nexus 1000v amd Switch Support-Virtual Vmware DVvs Scalability Open vSwitch 1000 hosts 1 Million hosts Cisco, Vmware, Arista, Brocade, Citrix, Red hat and Broadcom Arista 7150 and Brocade ADX Microsoft, Arista, Emulex, Huawei and HP No Vendor Support in Hardware Vmware and Broadcom Cisco No No

Fig 1.3- Comparison: VXLAN, NVGRE, STT and LISP