Scalability and Bandwidth Optimization for Data Center Networks

Data centers now being built may contain tens of thousands hosts, and this number is still growing. Deficiencies of the conventional network architecture are gradually revealed. It is facing a set of limitations such as poor scalability, resources fragmentation, lack of support for virtual machine migration and inefficient packets forwarding. In this paper, we propose an alternative solution which is more suitable for data center than the traditional structure. We redesign a large and scalable layer 2 network that supports VM migration better and removes resource fragmentation. We employ a global controller to maintain network information which can be accessed through a directory service running on the controller. With the directory service we achieve proxy ARP resolving and MAC-in-MAC tunneling while forwarding data frames. Moreover, we replace the shortest path first route algorithm with bandwidth-aware forwarding algorithm that dynamically adjusts route when overloaded link detected. The simulation results show that our approach can significantly increase the aggregate throughput and is more efficient.