Abstract Based on earlier work -xb[ 1 ] [ 2 ] [ 3 ], this paper presents a reformulation of network design (routing), replacing the traffic matrix, by a set of linear constraints, representing a convex polyhedral traffic ensemble. Linear programming is used, together with classical design techniques, to globally optimize a variety of metries over this traffic ensemble. Where applicable, the method yields richer answers than using a traffic matrix. Firstly, network costs can be globally bounded over correlated large scale variations of traffic, eliminating one of central weaknesses of classical network design (at least for planning purposes) — most traffic matrix forecasts are extremely error-prone. Secondly, inverse problems can be solved, with max/min bounds on network capacity and other metrics being derived from a prespecified cost. Thirdly, analyses of traffic matrices can be perfomed, detecting hidden biases in selection. Many other kinds of results can be derived, by appropriate reformulation of the governing LP. To the best of our knowledge, we present the first global bounds on network cost, with very limited knowledge of the traffic — total traffic, mean traversed distance and add-drop at nodes, are sufficient to obtain useful bounds on total network cost. The sufficient set of parameters is often less than 5% of the elements of the entire traffic matrix T and can be specified far more confidently than T . Our work has the potential to evaluate networks based on intrinsic topological and gross easily specified properties of the traffic , as opposed to point evaluations possibly modulated by uncertain traffic matrix data.
[1]
P. Harshavardhana,et al.
Broadband network infrastructure of the future: roles of network design tools in technology deployment strategies
,
1998,
IEEE Commun. Mag..
[2]
G.N.S. Prasanna.
Robust design of wireless backhaul networks using convex polyhedral representations of uncertainty
,
2002,
2002 IEEE International Conference on Personal Wireless Communications.
[3]
San-qi Li,et al.
Unified measurement functions for traffic aggregation and link capacity assessment
,
1999,
IEEE INFOCOM '99. Conference on Computer Communications. Proceedings. Eighteenth Annual Joint Conference of the IEEE Computer and Communications Societies. The Future is Now (Cat. No.99CH36320).
[4]
G. Prasanna,et al.
A reformulation of network design: replacing the uncertain traffic matrix, by a web of linear relations, and its implications
,
2003,
OFC 2003 Optical Fiber Communications Conference, 2003..