Large Layer 2 Network

Dynamic VM migration becomes a critical issue to meet flexible service changes. Dynamic VM migration is a process of moving VMs from one physical server to another, while ensuring normal running of the VMs. End users are unaware of this process, so administrators can flexibly allocate server resources or maintain and upgrade servers without affecting server usage by end users. The key of dynamic VM migration is to ensure uninterrupted services during the migration, so the IP and MAC addresses of VMs must remain unchanged. To meet this requirement, VM migration must occur within a Layer 2 domain but not across Layer 2 domains, as shown in Figure 1.

Figure 1 VM migration on a traditional network

In the traditional data center network architecture, the Layer 2 network uses redundant devices and links to improve reliability. This will inevitably result in physical loops during VM migration.

To prevent broadcast storms caused by physical loops, a loop prevention protocol such as Spanning Tree Protocol (STP) is required to block redundant links. Due to STP limitations, an STP-enabled Layer 2 network can contain no more than 50 network nodes, so dynamic VM migration can only occur in a limited scope.

To enable VM migration in a large scope or across domains, servers involved must be on the same Layer 2 network, which is called large Layer 2 network.

Network device virtualization, TRILL, and EVN technologies can construct a physical large Layer 2 network to enlarge the VM migration scope. However, a physical large Layer 2 network requires huge changes to the existing network structure, and still has many restrictions on the VM migration scope. VXLAN can solve the preceding problems.

A virtual large Layer 2 network can solve the problem and enable VM migration in a larger scope, as shown in Figure 2.

Figure 2 VM migration on a large Layer 2 network