Data centers play an important role in today’s Internet development. Research to find scalable architecture and efficient routing algorithms for data center networks has gained popularity. The fat-tree architecture, which is essentially a folded version of a Clos network, has proved to be readily implementable and is scalable. In this paper, we investigate routing on a fat-tree network by deriving its <italic>global packing number</italic> and by presenting explicit algorithms for the construction of optimal, load-balanced routing solutions. Consider an optical network that employs wavelength division multiplexing in which every user node sets up a connection with every other user node. The global packing number is basically the number of wavelengths required by the network to support such a traffic load, under the restriction that each source-to-destination connection is assigned a wavelength that remains constant in the network. In mathematical terms, consider a bidirectional, simple graph, <inline-formula> <tex-math notation="LaTeX">$G$ </tex-math></inline-formula> and let <inline-formula> <tex-math notation="LaTeX">$N\subseteq V(G)$ </tex-math></inline-formula> be a set of nodes. A path system <inline-formula> <tex-math notation="LaTeX">$\mathcal {P}$ </tex-math></inline-formula> of <inline-formula> <tex-math notation="LaTeX">$G$ </tex-math></inline-formula> with respect to <inline-formula> <tex-math notation="LaTeX">$N$ </tex-math></inline-formula> consists of <inline-formula> <tex-math notation="LaTeX">$|N|(|N|-1)$ </tex-math></inline-formula> directed paths, one path to connect each of the source-destination node pairs in <inline-formula> <tex-math notation="LaTeX">$N$ </tex-math></inline-formula>. The global packing number of a path system <inline-formula> <tex-math notation="LaTeX">$\mathcal {P}$ </tex-math></inline-formula>, denoted by <inline-formula> <tex-math notation="LaTeX">$\Phi (G,N,\mathcal {P})$ </tex-math></inline-formula>, is the minimum integer <inline-formula> <tex-math notation="LaTeX">$k$ </tex-math></inline-formula> to guarantee the existence of a mapping <inline-formula> <tex-math notation="LaTeX">$\phi :\mathcal {P}\to \{1,2,\ldots, k\}$ </tex-math></inline-formula>, such that <inline-formula> <tex-math notation="LaTeX">$\phi (P)\neq \phi (\widehat {P})$ </tex-math></inline-formula> if <inline-formula> <tex-math notation="LaTeX">$P$ </tex-math></inline-formula> and <inline-formula> <tex-math notation="LaTeX">$\widehat {P}$ </tex-math></inline-formula> have common arc(s). The global packing number of <inline-formula> <tex-math notation="LaTeX">$(G,N)$ </tex-math></inline-formula>, denoted by <inline-formula> <tex-math notation="LaTeX">$\Phi (G,N)$ </tex-math></inline-formula>, is defined to be the minimum <inline-formula> <tex-math notation="LaTeX">$\Phi (G,N,\mathcal {P})$ </tex-math></inline-formula> among all possible path systems <inline-formula> <tex-math notation="LaTeX">$\mathcal {P}$ </tex-math></inline-formula>. In additional to wavelength division optical networks, this number also carries significance for networks employing time division multiple access. In this paper, we compute by explicit route construction the global packing number of <inline-formula> <tex-math notation="LaTeX">$(\text {T}_{n},N)$ </tex-math></inline-formula>, where T<sub><italic>n</italic></sub> denotes the topology of the <inline-formula> <tex-math notation="LaTeX">$n$ </tex-math></inline-formula>-ary fat-tree network, and <inline-formula> <tex-math notation="LaTeX">$N$ </tex-math></inline-formula> is considered to be the set of all edge switches or the set of all supported hosts. We show that the constructed routes are load-balanced and require minimal link capacity at all network links.
[1]
D. West.
Introduction to Graph Theory
,
1995
.
[2]
Wing Shing Wong,et al.
The Global Packing Number for an Optical Network
,
2015,
ArXiv.
[3]
Gary Chartrand,et al.
The Colour Numbers of Complete Graphs
,
1967
.
[4]
Devavrat Shah,et al.
Fastpass: a centralized "zero-queue" datacenter network
,
2015,
SIGCOMM 2015.
[5]
Amin Vahdat,et al.
A scalable, commodity data center network architecture
,
2008,
SIGCOMM '08.
[6]
Ivan P. Kaminow,et al.
A Precompetitive Consortium on Wide-band All Optical Networks
,
1993
.
[7]
Gordon T. Wilfong.
Minimum Wavelength in an All-Optical Ring Network
,
1996,
ISAAC.
[8]
Jou-Ming Chang,et al.
Parallel construction of optimal independent spanning trees on hypercubes
,
2007,
Parallel Comput..
[9]
Stéphane Pérennes,et al.
Colouring Paths in Directed Symmetric Trees with Applications to WDM Routing
,
1997,
ICALP.
[10]
N. J. A. Sloane,et al.
The On-Line Encyclopedia of Integer Sequences
,
2003,
Electron. J. Comb..
[11]
Amin Vahdat,et al.
Jupiter Rising: A Decade of Clos Topologies and Centralized Control in Google's Datacenter Network
,
2015,
Comput. Commun. Rev..
[12]
Kaigui Bian,et al.
Control Channel Establishment in Cognitive Radio Networks using Channel Hopping
,
2011,
IEEE Journal on Selected Areas in Communications.
[13]
Charles Clos,et al.
A study of non-blocking switching networks
,
1953
.
[14]
Albert Y. Zomaya,et al.
Quantitative comparisons of the state‐of‐the‐art data center architectures
,
2013,
Concurr. Comput. Pract. Exp..
[15]
Ilyong Chung,et al.
Application of the Special Latin Square to a Parallel Routing Algorithm on a Recursive Circulant Network
,
1998,
Inf. Process. Lett..
[16]
Jonathan S. Turner,et al.
Multirate Clos networks
,
2003,
IEEE Commun. Mag..
[17]
Hung Q. Ngo,et al.
Multirate rearrangeable clos networks and a generalized edge coloring problem on bipartite graphs
,
2003,
SODA '03.
[18]
José R. Correa,et al.
Improved Bounds on Nonblocking 3-Stage Clos Networks
,
2007,
SIAM J. Comput..
[19]
Charles E. Leiserson,et al.
Fat-trees: Universal networks for hardware-efficient supercomputing
,
1985,
IEEE Transactions on Computers.
[20]
L. Gargano.
Colouring all directed paths in a symmetric tree with applications to WDM routing
,
1997
.
[21]
A. Mullin,et al.
Mathematical Theory of Connecting Networks and Telephone Traffic.
,
1966
.
[22]
Adel A. M. Saleh,et al.
All-Optical Networking—Evolution, Benefits, Challenges, and Future Vision
,
2012,
Proceedings of the IEEE.
[23]
P. Hell,et al.
Graph Problems Arising from Wavelength-Routing in All-Optical Networks
,
2004
.
[24]
Heiko Schröder,et al.
Optical All-to-All Communication for Some Product Graphs
,
1997,
SOFSEM.
[25]
Jonathan S. Turner,et al.
Nonblocking Multirate Networks
,
1989,
SIAM J. Comput..
[26]
Bruno Beauquier,et al.
All-to-all communication for some wavelength-routed all-optical networks
,
1999,
Networks.
[27]
J. Dénes,et al.
Latin squares and their applications
,
1974
.