Optimizing Data Centers with Large Layer 2 Network

In modern data center, large Layer 2 network play a crucial role in supporting high-performance and reliable networking for critical business applications. They simplify network management and enable adoption of new technologies, making them essential to data center architecture. This article will explore the necessity of a large Layer 2 network and the technologies used to implement them.

Why Is a Large Layer 2 Network Needed?

Traditional data center architecture typically follows a combination of Layer 2 (L2) and Layer 3 (L3) network designs, restricting the movement of servers across different Layer 2 domains. However, as data centers evolve from traditional setups to virtualized and cloud-based environments, the emergence of server virtualization technology demands the capability for dynamic VM migration. This process involves migrating a virtual machine from one physical server to another, ensuring it remains operational and unnoticed by end users. It enables administrators to flexibly allocate server resources or perform maintenance and upgrades on physical servers without disrupting users.

The key to dynamic VM migration is ensuring that services on the VM are uninterrupted during the transfer, which requires the VM’s IP address and operational state to remain unchanged. Therefore, dynamic VM migration can only occur within the same Layer 2 domain and not across different Layer 2 domains.

To achieve extensive or even cross-regional dynamic VM migration, all servers potentially involved in the migration must be included in the same Layer 2 domain, forming a larger Layer 2 network. This larger network allows for seamless, unrestricted VM migration across a wide area, known as a large Layer 2 network.

Large Layer 2 Network

How to Achieve a Truly Large Layer 2 Network?

The technologies for implementing large Layer 2 network can be divided into two main categories based on their source. One category is proposed by network equipment manufacturers, including network device virtualization and routing optimized Layer 2 forwarding technologies. The other category is proposed by IT manufacturers, including overlay technology and EVN technology.

Network Device Virtualization

Network device virtualization technology combines two or more physical network devices that are redundant with each other and virtualizes them into a logical network device, which is presented as only one node in the entire network. By combining network device virtualization with link aggregation technology, the original multi-node, multi-link structure can be transformed into a logical single-node, single-link structure. This eliminates the possibility of loops and removes the need for deploying loop prevention protocols. Consequently, the scale of the Layer 2 network is no longer constrained by these protocols, thereby achieving a large Layer 2 network.

Building a large Layer 2 network using network virtualization technology results in a logically simple network that is easy to manage and maintain. However, compared to other technologies, the network scale is relatively small. In addition, these technologies are the private technologies of each vendor, and can only use devices from the same vendor for networking, which is usually suitable for building large Layer 2 networks at the level of small and medium-sized PODs.

Routing Optimized Layer 2 Forwarding Technology

The core issue with traditional Layer 2 network is the loop problem. To address this, manufacturers insert additional headers in front of Layer 2 packets and use routing calculations to control data forwarding across the entire network. This approach extends the Layer 2 network’s scale to cover the entire network without being limited by the number of core switches, thereby achieving a large Layer 2 network.


The forwarding of Layer 2 messages by means of route computation requires the definition of new protocol mechanisms. These new protocols include TRILL, FabricPath, SPB, etc. Taking TRILL as an example, it transparently transmits the original Ethernet frame by encapsulating it with a TRILL header and a new outer Ethernet frame. TRILL switches forward packets using the Nickname in the TRILL header, which can be collected, synchronized, and updated through the IS-IS routing protocol. When VMs migrate within a TRILL network, IS-IS can automatically update the forwarding tables on each switch, maintaining the VM’s IP address and state, thus enabling dynamic migration.

TRILL enables the creation of larger Layer 2 network and, being an IETF standard protocol, simplifies vendor interoperability. This makes it ideal for large PODs or entire data centers. However, TRILL deployment often necessitates new hardware and software, which can result in higher investment costs.

Overlay Technology

Overlay technology involves encapsulating the original Layer 2 packets sent by the source host, transmitting them transparently through the existing network, and then decapsulating them at the destination to retrieve the original packets, which are then forwarded to the target host. This process achieves Layer 2 communication between hosts. By encapsulating and decapsulating packets, an additional large Layer 2 network is effectively overlaid on top of the existing physical network, so it is called overlay technology.

Overlay technology

This is equivalent to virtualizing the entire bearer network into a huge Layer 2 switch. Each virtual machine is directly connected to a port of this switch, so naturally there is no loop. The dynamic migration of a virtual machine is equivalent to changing the virtual machine from one port of the switch to another port, and the status can remain unchanged.

The overlay solution is proposed by IT vendors, such as VXLAN and NVGRE. In order to bulid an overlay network, FS has launched a VXLAN network solution, which uses VXLAN technology to fully improve network utilization and scalability. In the overlay solution, the bearer network only needs to meet the basic switching and forwarding capabilities, and the encapsulation and decapsulation of the original packets can be carried out by the virtual switches in the server, without relying on network devices.

EVN Technology

EVN (Easy Virtual Network) technology is designed for Layer 2 interconnection across data centers rather than within a single data center. Traditional methods like VPLS or enhanced VPLS over GRE often suffer from complex configurations, low bandwidth utilization, high deployment costs, and significant resource consumption. However, EVN, based on VXLAN tunnels, effectively addresses these issues and can be seen as an extension of VXLAN.

EVN technology uses the MP-BGP protocol to exchange MAC address information between Layer 2 networks and generates MAC address table entries for packet forwarding. It supports automatic VXLAN tunnel creation, multi-homing load balancing, BGP route reflection, and ARP caching. These features effectively address the issues found in VPLS and other Layer 2 interconnection technologies, making EVN an ideal solution for data center Layer 2 interconnection.


In this article, we discussed the importance of a large Layer 2 network in modern data centers, emphasizing its role in supporting virtualization, dynamic VM migrations, and the technologies needed for scalability. As an ICT company, FS is committed to being the top provider for businesses seeking dependable, cost-effective solutions for their network architecture. Utilizing our company’s advanced switches can significantly enhance the scalability of data centers, ensuring robust support for large Layer 2 networks. Register on our website today for more information and personalized recommendations.