 | This articleneeds additional citations forverification. Please helpimprove this article byadding citations to reliable sources. Unsourced material may be challenged and removed. Find sources: "Fat tree" – news ·newspapers ·books ·scholar ·JSTOR(August 2007) (Learn how and when to remove this message) |
Thefat tree network is a universalnetwork for provably efficient communication.[1] It was invented byCharles E. Leiserson of theMIT in 1985.[1] k-ary n-trees, the type of fat-trees commonly used in most high-performance networks, were initially formalized in 1997.[2]
In atreedata 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 atelecommunications network, the branches aredata links; the varied thickness (bandwidth) of the data links allows for more efficient and technology-specific use.[citation needed]
Mesh andhypercube topologies have communication requirements that follow a rigid algorithm, and cannot be tailored to specific packaging technologies.[3]
Supercomputers that use a fat tree network[4] include the two fastest as of late 2018,[5]Summit[6] andSierra,[7] as well asTianhe-2,[8] theMeiko Scientific CS-2,Yellowstone, theEarth Simulator, theCray X2, the Connection MachineCM-5, and variousAltix supercomputers.[citation needed]
Mercury Computer Systems applied a variant of the fat tree topology—thehypertree network—to theirmulticomputers.[citation needed] In this architecture, 2 to 360 compute nodes are arranged in acircuit-switched fat tree network.[citation needed] Each node has local memory that can be mapped by any other node.[vague] Each node in this heterogeneous system could be anIntel i860, aPowerPC, or a group of threeSHARCdigital signal processors.[citation needed]
The fat tree network was particularly well suited tofast Fourier transform computations, which customers used for suchsignal processing tasks asradar,sonar, andmedical imaging.[citation needed]
In August 2008, a team ofcomputer scientists atUCSD published a scalable design for network architecture[9] 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 aClos 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.
