Transcript Slide 1
CNP On Location of POPs in FTTH Networks using Center of Gravity Rasmus H. Nielsen Ph.D. Student Center for Network Planning (CNP) Department of Control Engineering Aalborg University, Denmark CNP Overview • • • • • • • • Motivation • Evolution of broadband services • Massive investments Optimization The facility location problem • Different norms Case study area Results Other methods Comparison Conclusion CNP Motivation – evolution of broadband services • ICT is becoming increasingly more important in society • Commercial companies and public institutions are becoming more and more dependent on ICT for every day operations. • Demand for higher bandwidth • Video conferences, backup, remote storage etc. • Demand for higher availability • E-commerce, sensitive data (tele medicine, control applications), communication with customers or other branches etc. • Increasing network usage and increasing expenses related to network downtime CNP Motivation – evolution of broadband services • The real driver is the convergence of medias and services, which is already increasing the demands to the network significantly CNP Motivation – massive investments • All over the world fiber is being deployed to fulfill the already existing or expected requirements • The last years have given us plenty of new fiber-related acronyms and terms to remember; FTTH, FTTC, VDSL, PON, Homerun, Triple Play etc. • In Asia and USA current telecommunication providers are taking part in the deployment of new ICT infrastructure • In Europe, and in Denmark in particular, the old telephony monopolies are hesitating to compete their own gold ore – the old copper networks CNP Motivation – massive investments • New players are entering the broadband market • Power companies have announced huge investments in fiber networks to a large part of the Danish households • Covering 40 % of all households • 1.3 billion Euros invested • Medio 2005 – 13,500 homes passed • Ultimo 2007 – 500,000 homes passed • With the current expected investments,1 million homes will be passed • Large investments –> large possible savings CNP Optimization of network deployment • Until recently all research within telephony networks were taking place within the telephony monopolies • With the deregulation of the telephony markets, many research labs were closed and the gathered knowledge became confidential • The knowledge of network deployment and optimization is limited • Many areas to consider: • Economics • Services • Management • Planning CNP Network planning • Considered in an application oriented scope – results should be put to commercial use – the sooner – the better • Any demand point (customer) must be served at a service point (POP) • The demand points are fixed – unfortunately customers will not move • Location of the service point is essential for the cost of lines and to some extent ducting CNP Facility location • Problem is well-considered within primarily operational research • • Different variants: P-median • Minimize line cost for a fixed number of demand points Uncapacitated • Minimize the cost of lines and demand points – find the most optimal combination Capacitated • As the uncapacitated but with an upper limit on the demand served at each service point Capacitated with Single Source • As the capacitated but with each demand point served completely from a service point • • • CNP Facility location – p-median • Optimize the location of a number of facilities (service points) in relation to a number of customers (demand points) • Brute force: n p 1 n! p!n p ! • The amount of data makes this infeasible. – n in the order of 10,000 – 100,000 • Heuristics are needed CNP P-median – center of gravity • • A simple first-try The service point should be located at the center of gravity of the demand points it services (here: that is closest to it) • Norms: xSP xDP 2 ySP yDP 2 • 1. Euclidean distance – • Cables will not be digged down as straight lines from service point to demand point • Computational fast • 2. Road distance – shortest path (Dijkstra) • More physical correct • Computational heavy CNP P-median • Distance matrix d1,1 D d n ,1 • di, j Cost C Yi , j Di , j j i • Connection matrix d1,m d n ,m y1,1 Y yn ,1 yi , j y1,m yn ,m Elements are 1 if the demand point is connected to the service point and 0 otherwise CNP Case Study Area • Municipality of Hals • Area – 191 km2 • Population – 11,500 • Population density • 60 citizens/km2 • Comparison [citizens/km2]: • Denmark: 126 • Greece: 84 • Athens: 19,619 CNP Results CNP Other methods for facility location • Tested so far: • Center of gravity • Very simple – Euclidean distance is very fast • Space filling curves • Another approach to density analysis • Genetic algorithms • Large potential and many possible extensions • Lagrangean relaxation • Well considered – guaranteed convergence CNP Lagrangean relaxation • Primary problem m n c x i 1 j 1 ij ij subject to m x i 1 ij 1 • Dual problem m cij xij s j 1 xij i 1 j 1 j 1 i 1 m n n subject to m y i 1 i p m xij yi i 1 yi , xij (0,1) yi p xij yi yi , xij (0,1) CNP Comparison p Center of Gravity Lagrangean Relaxation with subgradient optimization 4 2,477,938 2,136,171 -13.79 5 2,125,169 1,910,729 -10.09 6 1,824,115 1,822,845 -0.07 7 1,803,999 1,803,308 -0.04 8 1,840,690 1,823,764 -0.92 9 1,905,779 1,857,729 -1.52 10 1,970,287 1,894,557 -3.84 12 2,091,571 2,025,375 -3.16 14 2,191,479 2,170,971 -0.94 16 2,350,890 2,327,154 -1.01 18 2,523,818 2,487,093 -1.46 20 2,676,287 2,651,743 -0.92 CNP Conclusion • Good potential for significant savings • Optimization is needed now – in 10 years it will be too late • A 5% saving for the complete network is better than a 10% saving for 25% of the network • A method was tested and a better method has now been considered, which verify the results • Actually not that bad for a first try... • Even simple computer-aid easily beats the manual process when considering larger data sets CNP Further Research • Best method in sense of optimality and processing time • Include more parameters; • Cost – e.g. ducting • Market – e.g. penetration rate, demographic groups • Test methods on already planned areas – it is more exciting to compare to something “real”, than just comparing theoretical approaches • Integrate in a large framework of network planning optimizations CNP Thank you for your attention!