Two of the most popular data center networking architectures are the fat tree and jellyfish data center network architecture (DCNA). Being able to describe the differences in these two types of architectures will help you craft the perfect network for your enterprise. It is important to visually map out your network in order to catch any unforeseen problems that could arise. Once your ideal network is drawn out, you can begin interconnecting devices and enabling services for your end users.
The Fat Tree Network Architecture
An MIT professor by the name of Charles Leiserson is credited with founding the fat tree network architecture. The main premise of the fat tree network is that the links in the network design get bigger as packets move up into the networking tree. This method was designed to deliver bandwidth in an efficient, on-demand fashion. A fat tree network can be customized for specific data packaging techniques that can further tailor the network to the needs of the enterprise.
The basic premise of the internet operates on a fat tree network. In theory, your internet service provider links you into a fiber connection which is then linked to a network backbone which would be a broad example of the fat tree network in use. Keep in mind the internet uses peering therefore this modification has some experts calling the internet a modified or hybrid fat tree network.
The Jellyfish Data Center Networking Architecture
While the fat tree network architecture focuses on structure, the jellyfish network architecture decentralizes the network infrastructure in efforts to provide higher levels of bandwidth at the same cost of a fat tree network. The biggest questions newbies ask, “How is this possible?” At first glance, it could be easy to see why this question would be asked.
The jellyfish model seeks the most efficient route in order to send a packet; a fat tree design must adhere to the structure of the network that has been built out. This abstract method of networking is being used more and more in the cloud computing model because many computer scientists believe that data centers themselves could be linked together to operate on the jellyfish model.
This could provide end users with higher availability to enterprise level applications and more robust network resources while saving the data center administrators overhead on bandwidth and hardware costs associated with the fat tree design. Jellyfish’s creators mention that OpenFlow and MPLS assist in the routing traffic which has also helped bolster the implementation of this emerging architecture.