Template:Short description Template:More citations needed
The fat tree network is a universal network for provably efficient communication.<ref name="CL85">Template:Cite journal</ref> It was invented by Charles E. Leiserson of the MIT in 1985.<ref name="CL85" /> k-ary n-trees, the type of fat-trees commonly used in most high-performance networks, were initially formalized in 1997.<ref>Template:Cite book</ref>
In a tree data structure, every branch has the same thickness (bandwidth), regardless of their place in the hierarchy—they are all "skinny" (skinny in this context means low-bandwidth). In a fat tree, branches nearer the top of the hierarchy are "fatter" (thicker) than branches further down the hierarchy. In a telecommunications network, the branches are data links; the varied thickness (bandwidth) of the data links allows for more efficient and technology-specific use.Template:Citation needed
Mesh and hypercube topologies have communication requirements that follow a rigid algorithm, and cannot be tailored to specific packaging technologies.<ref>Template:Cite book</ref>
Applications in supercomputersEdit
Supercomputers that use a fat tree network<ref>Template:Cite book</ref> include the two fastest as of late 2018,<ref>{{#invoke:citation/CS1|citation |CitationClass=web }}</ref> Summit<ref>{{#invoke:citation/CS1|citation |CitationClass=web }}</ref> and Sierra,<ref>{{#invoke:citation/CS1|citation |CitationClass=web }}</ref> as well as Tianhe-2,<ref>{{#invoke:citation/CS1|citation |CitationClass=web }}</ref> the Meiko Scientific CS-2, Yellowstone, the Earth Simulator, the Cray X2, the Connection Machine CM-5, and various Altix supercomputers.Template:Citation needed
Mercury Computer Systems applied a variant of the fat tree topology—the hypertree network—to their multicomputers.Template:Citation needed In this architecture, 2 to 360 compute nodes are arranged in a circuit-switched fat tree network.Template:Citation needed Each node has local memory that can be mapped by any other node.Template:Vague Each node in this heterogeneous system could be an Intel i860, a PowerPC, or a group of three SHARC digital signal processors.Template:Citation needed
The fat tree network was particularly well suited to fast Fourier transform computations, which customers used for such signal processing tasks as radar, sonar, and medical imaging.Template:Citation needed
Related topologiesEdit
In August 2008, a team of computer scientists at UCSD published a scalable design for network architecture<ref name="FaLouVah08">Template:Cite book</ref> that uses a topology inspired by the fat tree topology to realize networks that scale better than those of previous hierarchical networks. The architecture uses commodity switches that are cheaper and more power-efficient than high-end modular data center switches.
This topology is actually a special instance of a Clos network, rather than a fat-tree as described above. That is because the edges near the root are emulated by many links to separate parents instead of a single high-capacity link to a single parent. However, many authors continue to use the term in this way.