Review of "VL2: A Scalable and Flexible Data Center Network"
VL2 is a practical network architecture that scales to support huge data centers with uniform high capacity between any servers, performance isolation between services and Ethernet layer-2 semantics. It uses flat addressing to allow service instances be put anywhere physically in the network, Valiant Load Balancing to spread traffic uniformly across network paths, and end-system based address resolution to scale to large server pools without complexing the network control plane. It puts an end to the expensive over-subscription model usually used to achieve high performance and is built on proven network technologies already available at a low cost in high-speed hardware implementations.
By offering layer-2 semantics, VL2 lets servers believe they share a single large IP subnet with other servers in the same service while eliminating the ARP and DHCP scaling bottlenecks. The network is built with two separate address families, topologically significant Locator Addresses (LA) and Application Addresses (AA). When a server S wants to send packet to server D, the VL2 agent on S traps the packets from S and wrap it with LA address of the ToR of the destination. Once the packet arrives at the destination ToR, the switch decapsulates the packet and deliver it to the destination AA carried in the inner header. Address resolution and access control is done in a similar way. The agent traps the request and query the directory service to get the corresponding information and then decide how to proceed.
Will this paper be influential in 10 years? Maybe. It proposes a good data center networking prototype that provides both good performance and flexibility. If it's also easy to use then it would be a good potential candidate for the next generation data center networking infra.