BMBF Project ROBUKOM: Robust Communication Networks

BMBF Project ROBUKOM: Robust Communication Networks Arie M.C.A. Koster Christoph Helmberg Andreas Bley Martin Grötschel Thomas Bauschert supported by BMBF grant 03MS616A: ROBUKOM Robust Communication Networks, www.robukom.de EuroView 2012 24 July 2012 Lehrstuhl II für Mathematik

Project ROBUKOM ROBUKOM - Robust Communication Networks joint BMBF project started in Oct 2010, until Jun 2013 11 academics in 5 research groups: 2 industrial partners: mathematical planning and operation of robust telecommunication networks different robustness concepts: classical survivability, demand uncertainty For further information, visit www.robukom.de Project 1 Robust Network Design under Demand Uncertainties RWTH Aachen: AK, Manuel Kutschka, Christina Büsing (associated) Project 2 Convex Chance-Constrained Models for Robust Communication Networks TU Chemnitz: Christoph Helmberg, Peter Hoffmann Project 3 Scalable Optimization Methods for Survivable IP-Networks TU Berlin: Andreas Bley, Daniel Karch Project 4 Efficient Algorithms for Survivable Multi-Layer Networks Zuse Institute Berlin: Martin Grötschel, Fabio D Andreagiovanni Project 5 Robust Design of IP-Networks with Multiple Transport-Technologies TU Chemnitz: Thomas Bauschert, Uwe Steglich Arie Koster RWTH Aachen University 2 / 17

Outline 1 Network Design under Demand Uncertainty 2 WDM Replacement Problem Arie Koster RWTH Aachen University 3 / 17

How robust is a network design? Demand uncertainties Traffic fluctuations in the US abilene Internet2 network in time intervals of 5 minutes during one week: Traffic fluctuates heavily between node-pairs Load of links will fluctuate alike Arie Koster RWTH Aachen University 4 / 17

Design of Telecommunication Networks discrete decisions: given a potential network topology, demand forecast, link modules with capacities/costs find a hardware configuration, and a routing, such that demands are satisfied and total installation cost is minimised. Variations / Extensions: Single path routing Integer routing Survivability requirements Node hardware (switching capacity) Wavelength assignment Multi-layer scenarios Arie Koster RWTH Aachen University 5 / 17

Deterministic Network Design f k ij = fraction of demand k K routed along arc (i, j) A. x e = number of link capacity modules to be installed on link e E. Arie Koster RWTH Aachen University 6 / 17

Deterministic Network Design f k ij = fraction of demand k K routed along arc (i, j) A. x e = number of link capacity modules to be installed on link e E. Integer Linear Programming formulation: min e E κ e x e s.t. {i,j} δ(i) k K (fij k fji k ) = (1) 1 i = s(k) 1 i = t(k), i V, k K (2) 0 else d k f k e Cx e, e E (3) f, x 0 (4) x Z E (5) Arie Koster RWTH Aachen University 6 / 17

Uncertain Demands Demand values d k are uncertain, i.e., random variables Arie Koster RWTH Aachen University 7 / 17

Uncertain Demands Demand values d k are uncertain, i.e., random variables Sufficient capacity cannot be guaranteed anymore! Arie Koster RWTH Aachen University 7 / 17

Uncertain Demands Demand values d k are uncertain, i.e., random variables Sufficient capacity cannot be guaranteed anymore! Capacity constraint (3) for edge e E changes to ( ) P d k fe k Cx e 1 ɛ, k K a so-called chance-constraint. Arie Koster RWTH Aachen University 7 / 17

Γ-Robust Network Design Approach by Bertsimas and Sim (2003,2004): d k [0, d k + ˆd k ], with nominal value d k and deviation ˆd k Arie Koster RWTH Aachen University 8 / 17

Γ-Robust Network Design Approach by Bertsimas and Sim (2003,2004): d k [0, d k + ˆd k ], with nominal value d k and deviation ˆd k Parameter Γ stating number of simultaneous deviations from nominal values For each deviating demand, the peak value d k + ˆd k describes the worst-case (capacity-wise) Arie Koster RWTH Aachen University 8 / 17

Computations Network Abilene Géant Germany17 Germany50 # nodes 12 22 17 50 # links 15 36 26 89 # demands 66 231 136 1044 available traffic period 6 months 4 months 1 day 1 day traffic granularity 5 min 15 min 5 min 5 min # available traffic matrices 48 095 10 737 288 288 # traffic matrices used 2 8 064 2 2 688 288 288 instances Abilene1 Geant1 Germany17 Germany50 Abilene2 Geant2 Arie Koster RWTH Aachen University 9 / 17

Gain of Robustness = Cost savings compared with 95% deterministic model Price of Robustness Arie Koster RWTH Aachen University 10 / 17

Gain of Robustness = Cost savings compared with 95% deterministic model Price of Robustness How robust are the network designs? Evaluation of 8064 traffic matrices (Abilene1, 4 weeks) realized robustness = maximal fraction of the total demand that can be routed for a traffic matrix Arie Koster RWTH Aachen University 10 / 17

Realized Robustness Abilene network: Evaluation of 8064 traffic matrices (4 weeks) Géant network: Evaluation of 2688 traffic matrices (4 weeks) Γ ABILENE1 ABILENE2 GEANT1 GEANT2 0 1 5 10 K Arie Koster RWTH Aachen University 11 / 17

Outline 1 Network Design under Demand Uncertainty 2 WDM Replacement Problem Arie Koster RWTH Aachen University 12 / 17

WDM replacement WDM upgrade procedure (1) Install new OXCs at node (In parallel to old one) (2) Shut down fiber(s) Replace regenerators Connect fiber(s) to new OXCs (3) Remove old OXCs from node Workforce constraint Step (2) requires personel at nodes and regenerators. Arie Koster RWTH Aachen University 13 / 17

Upgrading an Existing Network Existing connections are disrupted (a) Upgrade all fibers at once Uncritical: Only short break. Arie Koster RWTH Aachen University 14 / 17

Upgrading an Existing Network Existing connections are disrupted (a) Upgrade all fibers at once Uncritical: Only short break. (b) Upgrade fibers non-simultaneously Critical: Connection down until all fibers are upgraded. Why: Old & new WDM systems incompatible No mixed new-old connectons Arie Koster RWTH Aachen University 14 / 17

Effect of Disruption on Different Services IP service Down if IP network disconnected Arie Koster RWTH Aachen University 15 / 17

Effect of Disruption on Different Services IP service Down if IP network disconnected Unprotected WDM connection Down if (single) path disrupted Arie Koster RWTH Aachen University 15 / 17

Effect of Disruption on Different Services IP service Down if IP network disconnected Unprotected WDM connection Down if (single) path disrupted Protected WDM connection (1+1) Down if both paths disrupted Unprotected if one path disrupted Planning task Schedule fiber upgrades such that (1) IP service is never down (2) total WDM down time min (3) total WDM unprot. time min Arie Koster RWTH Aachen University 15 / 17

Conclusions Robust Network Design New ways to model network planning problem, allowing more information Mathematical Models are challenging generalization of existing theory Promising results: cheaper and better network designs Non-linear optimization approaches; techonology-dependent models WDM Replacement Problem New Problem many dependencies in solutions Straightforward formulation extremely weak First improvement results work in progress Arie Koster RWTH Aachen University 16 / 17