Analysis and development of data center networks

OData support
Supervisor:
Dr. Gulyás András
Department of Telecommunications and Media Informatics

The IT infrastructure is fundamental for a modern company. To ensure the operation of these companies fast and fault-tolerant IT systems are needed. The IT needs of the biggest companies can only be grant by special data centers. My task was to examine the existing data center structures and then plan a new data center where a greedy routing algorithm can be efficiently used.

The recently proposed data centers use special symmetric topologies. The reason for this is that on these topologies special routing algorithms can work, which can use the symmetry to provide short paths. The problem with these symmetric topologies is that they only let us to create specified size data centers, and these data centers can be hardly extended. I propose a special data center network which addresses this problem. I use hyperbolic tessellation in the proposed data center network, and in this work I show the properties of different tessellations. The different tessellations have different benefits, but they have one common property: any size of data center can be built using them, and extending these data centers is not a problem.

In the proposed data center I use a special geographic routing which is based on greedy forwarding. The routing is completely distributed, and can be used in any size of network. I show that the routing efficiently finds the shortest paths on the suggested topologies.

I also show that we don’t have to sacrifice the positive characteristics of the other data center networks if we choose the proposed data center network.

Downloads

Please sign in to download the files of this thesis.