Why Should We Integrate Services, Servers, and Networking In A Data Center?

WREN '09: Proceedings of the 1st ACM workshop on Research on enterprise networking |

Published by ACM

Publication

Since the early days of networks, a basic principle has been that endpoints treat the network as a black box. An endpoint injects a packet with a destination address and the network delivers the packet. This principle has served us well, and allowed us to scale the Internet to billions of devices using networks owned by competing companies and devices owned by billions of individuals. However, this approach might not be optimal for large-scale Internet data centers (DCs), such as those run by Amazon, Google, Microsoft and Yahoo, that employ custom software and customized hardware to increase efficiency and to lower costs. In DCs, all the components are controlled by a single entity, and creating services for the DC that treat the network as a black box will lead to inefficiencies.

In DCs, there is the opportunity to rethink the relationship between servers, services and the network. We believe that, in order to enable more efficient intra-DC services, we should close the gap between the network, services and the servers. To this end, we have been building a direct server-to-server network topology, and have been looking at whether this makes common services quicker to implement and more efficient to operate.