Image: Sumitomo Electric Lightwave Computer aided network planning and techno-economic assessment of next generation optical access networks Complex network planning methodology and framework Mitcsenkov Attila (BME TMIT, High Speed Networks Laboratory) 21 October 2014
Next Generation Access (NGA) networks • Optical fiber “approaches” the customers: - FTTC FTTB FTTH • A.) Point-to-point Dedicated optical fiber to all customers • B.) Point-multipoint systems Shared optical “feeder network” - Active (e.g. Active Ethernet) - Passive (PON)
Passive Optical Networks (PON) • „ History ”: ABON, BPON • Currently deployed: - EPON (USA & Asia) / GPON (EU) • Recent „upgrade” : - NGPON-1 10G EPON / 10G GPON • Near future: - NGPON-2 TWDM PON, 40G • Future development: - WDM PON ( “virtual point -to- point” ) - Colorless vs. Tunable: no clear winner yet…
Network design – The Motivation • Next Generation Access networks - Short / near term demand: 100 Mb/s – 1 Gb/s per customer - The common phrase: optical access is a “must” • „FTTH boom”: - Global Industry Analysts Inc.: 183.9 million FTTH/B subscriptions worldwide by 2015 - Deutche Telekom “will invest almost 30 billion euros over the next three years in order to ensure “the future of telecommunications.” (2012 – FTTC VDSL, Fiber & LTE) - … • What does it cost? • Who will pay for it? • Does it make sense?
The “crystal ball” : Techno-economic analysis • Economic aspects: - Cost estimation - Return of Investment, Cash Flow, … Business Case Analysis - “What do engineers now about business…”” • Technical aspects: - Quality parameters ( bandwidth, delay, reliability, … ) - Support network planning: what and where to deploy? - “Never let the marketing guys decide…” • Decision support: - Lowest cost vs. future proof technologies (e.g. VDSL vs. WDM PON) - Best fitting solutions for each service area
The beginning Once upon a time… • Some fundamental questions: - What is the cost of deploying a GPON network in XY area? - And what about Active Ethernet, VDSL or Point-to-Point? - Which one is the best? - ( what does “best” mean ?) ? ? ?
„ Think, think, think …” • How to estimate the costs? • Who will trust our estimation? • Why exactly *** € / $? Why not less/more? • We did a research of the literature (not to reinvent the wheel…) - Rules of thumb : average cost per customer (but we didn’t have data for Hungary…) - Geometric models to estimate the physical infrastructure (but our areas were not that regular / homogeneous…)
Geometric models 1. Triangle Model A FP1 FP2 végpont α Kábel típus #1 Kábel típus #2 B Kábel típus #3 R Kábel típus #4 F F Kábel típus #5 F F C Analytic calculations F F F D E F AB BC R / 3 cos( / 2) F F F 2 CD CE R / 6 1 8 sin ( / 2) DF R 0,132 0,336 n
Geometric models 2. Simplified Street Length model N: buildings in n houses in a row each street L: distance between l = distance neighboring buildings between two houses The central office CO – Central Office in the at the center geometric median, serving n 2 This serves n 2 customers customers Analytic calculations 2 I n (n 1) l (n 1) l (n 1) l n 1 F 4 l min i,n i n i i 1
Think, think, think …” • How to estimate the costs? • Who will trust our estimation? • Why exactly *** € / $? Why not less/more? • We did a research of the literature (not to reinvent the wheel…) • Investigate the map - “How big” network? - How complicated “field”? - How to adapt the geometric models? - Knowing the network, we could answer all questions…
Solution: Strategic design • High level network design - Network connections, system design, location of equipment, … • Detailed and accurate information about the necessary physical infrastructure - Network equipment - Cable plant - Optical fiber needs • Provide “Bill of Material” • Techno-economic analysis: - Reliable preliminary cost estimation - Comparison of technologies
How to provide the “strategic design”? • Design “guesswork” : • Computer aided network design: - Human creativity - Time and cost savings - Time consuming - Repetitious calculations with variable parameters - N-1 unnecessary designs! - Highly complex - Data availability - Computing capacity
Opportunities that made the computer able to deliver • Geographic Information System (GIS) data and digital maps • Increased computational capacity • Modern algorithms Map based, automatic design of access networks • The technical preconditions were not met earlier (calling for the use of geometric models)
Network planning as an optimization problem Challenges, model, cost function and constraints
What do we expect? What is a “good” network design?
What do we expect? What is a “good” network design?
Which one is the “better” PON network? Why?
Challenges • How to translate the problem to mathematics? - Modeling - Cost function - Constraints • What is an optimal, or at least good topology? - Economic aspects: what costs to consider? • Complexity - NP-hard (@see Travelling Salesman Problem) - Large scale problem instances (10.000 + households)
Modeling – translate it to the computer! Point-to-multipoint networks Központ (Central Központ (Central co Office, CO) Office, CO) Központ (Central co Megengedett Végpont / Office, CO) elosztópont hely előfizető Megengedett e 2 e 2 e 2 elosztóponti hely Elosztópont Gráfpont Elosztópont co e 1 e 4 Végpont Végpont co e 1 e 1 e 4 e 4 Gráfél Törzshálózat Gráfpont Gráfpont CO e 3 e 3 e 3 Lehetséges Törzshálózati Elosztóhálózat hálózati összeköttetés összeköttetés Elosztóhálózati összeköttetés {i} Előfizetői Nem használt csoport hálózati szakasz Törzshálózat Elosztóhálózat
Network graph model • Flexible, accurate, parametric • Contains all necessary data – and nothing else! - Potential network links - Distances - Customer data - Equipment locations
Goal: minimize costs CAPEX OPEX Deployment cost Operation, Maintenance Topology-dependent costs (e.g. cable plant) Topology-independent costs (e.g. user modems)
Cost function „ All inclusive ” Cable plant - Two levels: deployment + fiber costs Kábelhálózat költsége - Stepwise: discrete increments C 0 +n · C v - Existing infrastructure (!) Szálköltség C 0 } Telepítési Network equipment költésg - Central Office Kábelhálózat létrehozása Építési költségek - Distribution Units # szálszám - Customers Relation between these: - More distribution units less cable/fiber infrastructure - Two extremities: point-to-point network (1 distribution units / ∞ distribution units)
Physical and technical constraints • Constraint set: - Distance limitations - Complete access network reach - Feeder network reach - Distribution network reach - Differential distance Központ - Capacity of distribution units Végpont Elosztópont (Pl.. ETH Switch) max Törzshálózat L dist {i} Elosztóhálózat {i} Előfizetői csoport Törzshálózat L feed L feed hatósugara max max Elosztóhálózat max L dist hatósugara Elosztóhálózat hatósugara Törzshálózat hatósugara
The challenge is given … How to solve?
1. Subproblems and decomposition 2. Specialization • Fundamentals of optimization: “ No Free Lunch” Theorem » Specialized algorithms needed - Follow the designers’ way of thinking - Identify technology specific constraints and costs
Passive Optical Networks (GPON) • Costs: - Cable plant dominates - Splitter costs: low Központ • Network reach: 20 km not GPON 10GPON Végpont decisive {i} (WDM-PON) Splitter • Primary goal: optimal clustering Törzshálózat Elosztóhálózat • Solution: {i} Előfizetői L max csoport - Maximize shared cable segments Hálózat L max hatósugara - Clustering on the Shortest Path Tree - Typically tree topologies …
Branch Contracting Algorithm
Active Optical Networks (AETH) • Costs: - Distribution Unit: high CAPEX + OPEX - Cable plant: high costs… • Network reach: 20+3 km (not sharp) Központ Active • Main goal: Végpont Ethernet Elosztópont - Minimize distribution units / groups (Pl.. ETH Switch) max L dist Törzshálózat {i} - Optimal clustering: Elosztóhálózat {i} Előfizetői - No overlapping csoport Törzshálózat - Nearby distribution unit L feed L feed hatósugara max max Elosztóhálózat max L dist hatósugara Elosztóhálózat hatósugara • Solution: Törzshálózat hatósugara - Iterative “bottom - up” clustering - Increase group size if possible - Avoid overlapping
Recommend
More recommend
