Southern Africa Telecommunication Networks and
Applications Conference (SATNAC) 2014
31 August - 3 September 2014
Boardwalk Conference Centre, Nelson Mandela Bay, Eastern Cape, South Africa
Ubiquitous Broadband – An enabler to transform lives
Venue:
Boardwalk Conference Centre, Nelson Mandela Bay,
Eastern Cape, South Africa
Date:
31 August to 3 September 2014
Publication Information
Title: Southern Africa Telecommunication Networks and Applications Conference (SATNAC) 2014 Proceedings Format: Printed
ISBN: 978-0-620-61965-3
Format: CD-ROM
Editor: Roy Volkwyn (Telkom)
Date of print: August 2014
Version: First Edition
SATNAC is the flagship of the Telkom Centre of Excellence (CoE) Programme.
Page iii
Southern Africa Telecommunication Networks and Applications Conference (SATNAC) 2014Table of Contents
SATNAC 2014 Conference Sponsors Page vii
SATNAC 2014 Review Process Page ix
Organizing Committee Page xi
Technical Programme Committee Page xi
Index Page xix
Full papers Page 1
SATNAC 2014 Conference Sponsors
The SATNAC 2014 Committee would like to recognize the following sponsors
:Diamond Sponsors
Platinum Sponsors
Gold Sponsors
SATNAC 2014 Review Process
A formal ‘Call for Papers’ was issued, inviting anyone interested to submit a paper within categories specified by the Organizing Committee. Authors uploaded their papers via web interface onto a database. Papers were assigned to the review panel in the field to judge on the possible acceptance of the submission, based on the scope and depth of the subject matter.
The review process is based on the international de facto standard for blind paper reviews. The review process was undertaken by at least three experienced and well respected individuals. In the blind peer-review process, papers were scrutinized by a panel of South African reviewers, consisting of mainly respected academics, as well as several international experts. The reviewers were asked to provide specific feedback, both positive and negative. This was the only information from the review process disclosed to the authors; all other information was kept confidential.
Reviewers used a 4 point scale to rate the following criteria: • Originality
• References • Technical Quality • Presentation Style
Reviewers gave an overall rating. This was followed by the reviewer comments, which assists the authors in improving and correcting their papers. Reviewers were asked to be as comprehensive as possible.
The reviewers submitted their scoring and comments via web interface onto the database. The Technical Programme Committee drew reports and aggregated the individual scores. The papers were ranked on their average weighted score. The programme dictated the number of papers that could be accepted. Papers were submitted to an online plagiarism database, before being accepted. The reviewers’ comments were forwarded to the author’s, with a request to submit a final revised version. Only those papers of high enough quality as recommended by the respective reviewers are included in the SATNAC 2014 Proceedings as Full Reviewed Papers.
Two page Work-In-Progress papers were also invited but were not reviewed as rigorously. Several were accepted for oral presentations, while others for poster presentations. The poster session papers do not form part of the official conference proceedings.
Roy Volkwyn Chairperson
Technical Programme Committee SATNAC 2014
Page ix
Southern Africa Telecommunication Networks and Applications Conference (SATNAC) 2014Organising Committee
Alphonzo Samuels (Chairperson) Gys Booysen
Marti Beukes Graeme Allan
Technical Programme Committee
Mr. Roy Volkwyn (TPC Chairperson), Telkom SA Mr. Gys Booysen, Telkom SA
Mr. Chris Chavaranis, Alcatel-Lucent Dr. Johann Kellerman, Alcatel-Lucent Ms. Merryl Ford, CSIR
Mr. Emmanuel Adigun, Deloitte Dr. I-Sheng Liu, Ericsson Mr. Nishkar Govender, Ericsson Ms. Venita Engelbrecht, FibreCo
Prof. Dr. Thomas Magedanz, Frauenhofer Institut
Dr. Anish Kurien, F'SATI/Tshwane University of Technology Prof. K Djouani, F'SATI/Tshwane University of Technology Prof TO Olwal, F'SATI/Tshwane University of Technology Prof. Y Hamam, F'SATI/Tshwane University of Technology Mr. Abri Rozendaal, IBM
Mr.Emile Swanson, IBM
Mr. Waheed Swales, Investment Solutions Limited Ms. Phillippa Wilson, Jasco
Dr. Rolan Christian, KPMG
Dr. Paul Plantinga, Monash University Prof. Andre Calitz, NMMU
Prof. Charmain Cilliers, NMMU Mr Clayton Burger, NMMU Prof. Janet Wesson, NMMU Prof. Jean Greyling, NMMU Dr. Lester Cowley, NMMU
Mr. Patrick Tchankue Sielinou, NMMU Ms. Simone Beets, NMMU
Prof. Albert Helberg, North West University Prof. Alwyn Hoffman, North West University Mr. Arno de Coning, North West University Dr. Charl van Heerden, North West University Gerhard de Klerk, North West University Prof. Hennie Kruger, North West University Mr. Henri Marais, North West University Ms. Leenta Grobler, North West University Dr. Melvin Ferreira, North West University Prof. Marelie Davel, North West University
Mr. Samuel van Loggerenberg, North West University Prof. SE Terblanche, North West University
Dr. Tiny du Toit, North West University Prof. Willie Venter, North West University Dr. James Whitehead, Reutech Communications
Page xi
Southern Africa Telecommunication Networks and Applications Conference (SATNAC) 2014Prof. Alfredo Terzoli, Rhodes University Dr. Barry Irwin, Rhodes University Prof. George Wells, Rhodes University Prof. G. Foster, Rhodes University
Prof. Hannah Thinyane, Rhodes University Ms. Ingrid Sieborger, Rhodes University Mr. James Connan, Rhodes University Dr. Karen Bradshaw, Rhodes University Mr. Kevin Duff, Rhodes University Prof. Mici Halse, Rhodes University Dr. Mosiuoa Tsietsi, Rhodes University Prof. Peter Wentworth, Rhodes University Prof. Philip Machanick, Rhodes University Prof. Richard Foss, Rhodes University Mr. Yusuf Motara, Rhodes University Dr. Richard Good, Smile Communications Dr. Thomas Niesler, Stellenbosch University Prof. Tony Krzesinski, Stellenbosch University Prof. Johan du Preez, Stellenbosch University Ms. Charna John, Telkom
Mr. David van der Merwe, Telkom Mr. Eddie Mamahlodi, Telkom Mr. Edward Lebese, Telkom Mr. Eric Cartwright, Telkom Mr. Grant Evert, Telkom Mr. Ian Durston, Telkom Dr. Imran Achmed, Telkom Mr. Jaco Venter, Telkom Ms. Mariska de Lange, Telkom Ms. Melanie Delport, Telkom Ms. Meredith van Rooyen, Telkom Ms. Mphakiseng Masuabi, Telkom Mr. Nigel Naidoo, Telkom
Mr. Per Klotzsch, Telkom Mr. Siphile Sibaya, Telkom Mr. Sherwin Barlow, Telkom Mr. Tebogo Modiba, Telkom Mr. Tennyson Chimbo, Telkom Mr. Theran Naidoo, Telkom Mr Zaid Paruk, Telkom
Ms. Zamandlela Ndlela, Telkom
Dr. L. Magagula, Tshwane University of Technology Mr. Abel Ajibesin, University of Cape Town
Mr. Akinyemi Lateef Adesola, University of Cape Town Dr. Alexandru Murgu, University of Cape Town
Dr. Boyan Soubachov, University of Cape Town Mr. Clifford Sibanda, University of Cape Town Prof. Edwin Blake, University of Cape Town
Mr. Enoruwa Obayiuwana, University of Cape Town Mr. Henry Ohize, University of Cape Town
Prof. Hussein Suleman, University of Cape Town Ms. Joyce Mwangama, University of Cape Town
Page xiii
Southern Africa Telecommunication Networks and Applications Conference (SATNAC) 2014Prof. Mqhele Dlodlo, University of Cape Town Mr. Neco Ventura, University of Cape Town Mr. Nicholas Katanekwa, University of Cape Town Dr. Olabisi Falowo, University of Cape Town
Mr. Periola Ayodele Abiola, University of Cape Town Mr. Richard Spiers, University of Cape Town
Mr. Samson Orimolade, University of Cape Town Prof Tania Douglas, University of Cape Town Dr. Khulumani Sibanda, University of Fort Hare Prof. Mamelo Thinyane, University of Fort Hare Mr. Mfundo Shakes Scott, University of Fort Hare Mr. Sikhumbuzo Ngwenya, University of Fort Hare Mr. Zelalem Shibeshi, University of Fort Hare Prof. Andre Nel, University of Johannesburg Dr. Khmaies Ouahada, University of Johannesburg Dr. Rodolfo Martinez, University of Johannesburg Dr. Meera Joseph, University of Johannesburg Mr. P Robinson, University of Johannesburg
Dr. Suvendi Chinnappen, University of Johannesburg Mr. Y Roodt, University of Johannesburg
Mr Bashan Naidoo, University of KwaZulu-Natal Prof. Hongjun Xu, University of KwaZulu-Natal Prof. Jules Tapamo, University of KwaZulu-Natal Dr. Narushan Pillay, University of KwaZulu-Natal Dr. Olutayo O. Oyerinde, University of KwaZulu-Natal Dr. P.A.Owolawi, University of KwaZulu-Natal Mr S Rezenom, University of KwaZulu-Natal Prof. S H Mneney, University of KwaZulu-Natal Dr. Tahmid Quazi, University of KwaZulu-Natal Prof. Thomas Afullo, University of KwaZulu-Natal
Prof. Viranjay M. Srivastava, University of KwaZulu-Natal Mr. Jonas Manamela, University of Limpopo
Prof. Attahiru Sule Alfa, University of Manitoba Mr Babatunde Awoyemi, University of Pretoria Mr. Bruno de Carvalho e Silva, University of Pretoria Mr. Colman Mbuya, University of Pretoria
Mr. Hans Grobler, University of Pretoria Mr. Jacques van Wyk, University of Pretoria Prof. Louis Linde, University of Pretoria Mr. Malcolm Sande, University of Pretoria Mr. Mike Asiyo, University of Pretoria
Mr. Pieter Jansen van Vuuren, University of Pretoria Dr. Reza Malekian, University of Pretoria
Mr Roy Fisher, University of Pretoria Mr. Simon Barnes, University of Pretoria Ms. Shruti Lall, University of Pretoria Mr Smart Lubobya, University of Pretoria Dr. Vivek Dwivedi, University of Pretoria Mr. Pheeha Machaka, University of South Africa Mr. Adiel Ismail, University of Western Cape Prof. Antoine Bagula, University of Western Cape Mr. Ian Cloete, University of Western Cape
Page xv
Southern Africa Telecommunication Networks and Applications Conference (SATNAC) 2014Prof. Isabella M. Venter, University of Western Cape Mr. Mehrdad Ghaziasgar, University of Western Cape Mr. Michael J. Norman, University of Western Cape Mr. Reginald M. Dodds, University of Western Cape Prof. William D. Tucker, University of Western Cape Mr. Bethel Mutanga, University of Zululand
Mr. Olukayode A. Oki, University of Zululand Mr. Paul Tarwireyi, University of Zululand Mr. Pragasen Mudali, University of Zululand Mr. Alaba Akingbesote, University of Zululand Prof. Christo Pienaar, Vaal University of Technology Mr. Familua Ayokunle Damilola, Witwatersrand University Prof. Fambirai Takawira, Witwatersrand University
Dr. Jaco Versfeld, Witwatersrand University Dr. Ling Cheng, Witwatersrand University
Mr. Muhammad Waqar Saeed, Witwatersrand University Mr. Radu-Ionut Constantinescu, Witwatersrand University Ms. Reevana Balmahoon, Witwatersrand University Dr. Renier Dreyer, Witwatersrand University Prof. Rex van Olst, Witwatersrand University Mr Stephen Chabalala, Witwatersrand University
Page xvii
Southern Africa Telecommunication Networks and Applications Conference (SATNAC) 2014SATNAC 2014 Technical Programme
1. Access Networks Technologies
Title: 26: Optimal Passive Optical Network Planning Under Demand Uncertainty Page 3
Authors: Samuel van Loggerenberg (North-West University), Melvin Ferreira (North-West University), Leenta Grobler (North-West University) and Fanie Terblanche (North-West University)
Title: 97: Performance Analysis of LDPC-Based IEEE 802.16e with Different Modulation Techniques Page 9 Authors: Idris Adedapo Abimbola (Tshwane University of Technology), Mjumo Mzyece (Tshwane University of
Technology) and Guillam Noel (Tshwane University of Technology)
Title: 22: Iterative zero-forcing MIMO Decoder with Symbol Sorting Page 15 Authors: Philip Botha (University of Pretoria) and Sunil Maharaj (University of Pretoria)
Title: 50: Experimental Demonstration of Raman Amplification in Vertical Cavity Surface Emitting
Lasers for Extended Reach Access Networks Page 21
Authors: Enoch Kirwa Rotich Kipnoo (Nelson Mandela Metropolitan University), Valentine Tichakunda Chabata (Nelson Mandela Metropolitan University), Romeo Gamatham (Nelson Mandela Metropolitan
University), Andrew Leitch (Nelson Mandela Metropolitan University) and Tim Gibbon (Nelson Mandela Metropolitan University)
Title: 102: Optimal Decoding of the Alamouti 4×2 Space-Time Block Coding Page 27 Authors: Witesyavwirwa Vianney Kambale (Tshwane University of Technology), Karim Djouani (Tshwane
University of Technology) and Anish Mathew Kurien (Tshwane University of Technology)
Title: 67: A Cross-layer Based Subchannel Allocation Scheme in Satellite LTE Networks Page 33 Authors: Gbolahan Aiyetoro (University of KwaZulu-Natal) and Fambirai Takawira (University of the
Witwatersrand)
Title: 34: CDMA-DCDM for Cognitive Radio Networks Page 39 Authors: Periola Ayodele (University of Cape Town) and Falowo Olabisi (University of Cape Town)
Title: 68: Design of a Cognitive Small Cell Backhaul System for Non-Line-of-Sight Deployment in
Urban Canyons Page 45
Authors: Bessie Malila (University of Cape Town), Olabisi Falowo (University of Cape Town) and Neco Ventura (University of Cape Town)
Title: 55: Optimisation of SlotTime for a single-radio Mid-Range Multi-hop Wireless Mesh Network Page 51 Authors: Carlos Rey-Moreno (University of the Western Cape), Willaim D. Tucker (University of the Western
Cape) and Javier Simó-Reigadas (University Rey Juan Carlos)
Title: 114: A Hybrid Fuzzy Logic-Based Call Admission Control in LTE Networks Page 57 Authors: Christophe Boris Tokpo Ovengalt (Tshwane University of Technology), Karim Djouani (Tshwane
University of Technology) and Anish M Kurien (Tshwane University of Technology)
Title: 49: Rainfall Cell Estimation and Attenuation Studies for Radio links at Subtropical Africa Page 61 Authors: Akintunde Ayodeji Alonge (University of KwaZulu-Natal) and Thomas Joachim Afullo (University of
KwaZulu-Natal)
Title: 10: Prediction of Time-series Rain Attenuation based on Rain Rate using Synthetic Storm
Techniques over a Subtropical Region Page 67
Authors: Joseph Sunday Ojo (Mangosuthu University of Technology) and Pius Adewale Owolawi (Mangosuthu University of Technology)
Title: 86: Multicast Group Flow Rate Scaling in WiMAX Networks Page 73 Authors: Didacienne Mukanyiligira (University of Cape Town) and Alexandru Murgu (University of Cape Town)
Page xix
Southern Africa Telecommunication Networks and Applications Conference (SATNAC) 20142. Converged Services
Title: 99: Integration of Phonotactic Features for Language Identification on Code-Switched Speech Page 81
Authors: Koena Ronny Mabokela (University of Limpopo) and Madimetja Jonas Manamela (University of Limpopo)
Title: 98: Rendering South African Sign Language sentences from SignWriting notation Page 87
Authors: Kenzo Abrahams (University of the Western Cape), Mehrdad Ghaziasgar (University of the Western Cape), James Connan (Rhodes University) and Reg Dodds (University of the Western Cape)
Title: 37: Graphemes and Phonemes as Acoustic Sub-word Units for Continuous Speech Recognition
of Under-resourced Languages Page 93
Authors: Mabu Johannes Manaileng (University of Limpopo) and Madimetja Jonas Manamela (University of Limpopo)
Title: 92: Towards Development of A Stemmer for the IsiXhosa Language Page 99
Authors: Mnoneleli Nogwina (University of Fort Hare), Zelalem Shibeshi (University of Fort Hare) and Zoliswa Mali (University of Fort Hare)
Title: 28: Digital Video Shot Boundary Detector Investigation Page 105
Authors: M.G. De Klerk West University), W.C. Venter West University) and A.J. Hoffman (North-West University)
Title: 107: A Model for Context Awareness for Mobile Applications using Multiple-Input Sources Page 111 Authors: Direshin Pather (Nelson Mandela Metropolitan University), Janet Wesson (Nelson Mandela
Metropolitan University) and Lester Cowley (Nelson Mandela Metropolitan University)
Title: 31: Optical Character Recognition Using Minutiae Based Feature Detection Page 117 Authors: Pieter Erasmus (Hypervision Research Laboratory), Trevor Ho (Hypervision Research Laboratory) and
Yuko Roodt (Hypervision Research Laboratory)
Title: 41: HADEDA: A Concurrent Music Synthesis Project for the XMOS startKIT Page 123 Authors: James Dibley (Rhodes University) and Karen Bradshaw (Rhodes University)
Title: 5 111: Development of Soundex Algorithm for IsiXhosa Language
Authors: Zukile Ndyalivana (University of Fort Hare) and Zelalem Shibeshi (University of Fort Hare) Page 129
3. Core Network Technologies
Title: 5 2: Implementation of EPC Mobile Networks using NFV and SDN
Authors: Joyce Mwangama (University of Cape Town) and Neco Ventura (University of Cape Town) Page 137
Title: 82: Design of a Network Packet Processing Platform
Authors: Sean Pennefather (Rhodes University) and Barry Irwin (Rhodes University) Page 143
Title: 4: An Approach to Providing Quality of Service (QoS) for Over the Top (OTT) Voice in LTE
Networks
Authors: Nikesh Nageshar (University of the Witwatersrand) and Rex van Olst (University of the Witwatersrand) Page 149
Title: 51: A Performance Analysis of the Phase Shift and Pulse Delay Techniques for Chromatic
Dispersion Measurements and Compensation in Single Mode Fibre
Authors: Shukree Wassin (Nelson Mandela Metropolitan University), Enoch Kirwa Rotich Kipnoo (Nelson Mandela Metropolitan University), Romeo Gamatham, Andrew Leitch (Nelson Mandela Metropolitan University) and Tim Gibbon (Nelson Mandela Metropolitan University)
Page 155
Title: 52: Analysis of Optical Signal to Noise Ratio in Modern Transmission Fibres during Raman
Amplification
Authors: George Isoe (University of Eldoret), Kennedy Muguro (University of Eldoret), David Waswa (University of Eldoret), Enoch Rotich Kipnoo (Nelson Mandela Metropolitan University), Tim Gibbon (Nelson Mandela Metropolitan University) and Andrew Leitch (Nelson Mandela Metropolitan University)
Page 161
4. Internet Services & End User Applications
Title: 63: Designing Novel Visualisation Techniques for Managing Personal Information across
Multiple Devices
Authors: Simone Beets (Nelson Mandela Metropolitan University) and Janet Wesson (Nelson Mandela
Metropolitan University) Page 167
Title: 89: A Comparison of Machine Learning Techniques for Hand Shape Recognition
Authors: Roland G. Foster (University of the Western Cape), Mehrdad Ghaziasgar (University of the Western Cape), James Connan (University of the Western Cape) and Reg Dodds (University of the Western Cape)
Page 173
Title: 71: Spam Email Classification with Generalized Additive Neural Networks using Ensemble
Methods
Authors: Pieter Labuschagne (North-West University) and Tiny Du Toit (North-West University) Page 179
Title: 5 15: Development of an online reputation monitor
Authors: Gert Venter West University), Willie Venter West University) and Alwyn Hoffman
(North-West University) Page 185
Title: 85: Facebook Crawler Architecture for Opinion Monitoring and Trend Analysis Purposes
Authors: Sinesihle Ignetious Mfenyana (University of Fort Hare), Nyalleng Moorosi (University of Fort Hare) and
Mamello Thinyane (University of Fort Hare) Page 191
Title: 20: Online assignment Submissions at an ODL institute – Revelations of Current Internet
Traffic
Authors: Arthur James Swart (Central University of Technology) Page 197
Title: 5 72: Securing Mobile Payments on Unsecure Mobile Devices
Authors: Rossouw de Bruin (University of Johannesburg) and Sebastian von Solms (University of
Johannesburg) Page 203
Title: 5 61: Mobile Health Monitoring System for Community Health Workers
Authors: George Sibiya (Council for Scientific and Industrial Research), Ishmael Makitla (Council for Scientific and Industrial Research), Samuel Ogunleye (Council for Scientific and Industrial Research), Thomas Fogwill and Ronell Aberts (Council for Scientific and Industrial Research)
Page 209
Title: 3: A Consumer Health Informatics Application for e-Health Interventions in Marginalised Rural
Areas
Authors: Chikumbutso Gremu (Rhodes University), Alfredo Terzoli (Rhodes University) and Mosiuoa Tsietsi
(Rhodes University) Page 215
Title: 5 93: Contract-based Web Service Evolution Model
Authors: Kudzai Chiponga (University of Zululand), Paul Tarwireyi (University of Zululand) and Matthew Adigun
(University of Zululand) Page 221
Title: 5 121: Using a Mobile Solution to Support Chronic Disease Management in South Africa
Authors: Cainos Mukandatsama (Nelson Mandela Metropolitan University) and Janet Wesson (Nelson Mandela
Metropolitan University) Page 227
Title: 5 110: SerPro: a Mashup Tool for Enhanced Usability
Authors: Sabelo Yalezo (University of Fort Hare) and Mamello Thinyane (University of Fort Hare) Page 233
Title: 62: Transforming Learning: a Web-based M-learning System for Ad-hoc Learning of
Mathematical Concepts Amongst First Year Students at the University of Namibia
Authors: Ndapewa Ntinda (Rhodes University), Hannah Thinyane (Rhodes University) and Ingrid Sieborger
(Rhodes University) Page 239
Page xxi
Southern Africa Telecommunication Networks and Applications Conference (SATNAC) 20145. Limited Range Communications
Title: 91: Comparison of Energy-based Leader Selection Algorithms in Wireless Mesh Networks
Authors: Olukayode Oki (University of Zululand), Pragasen Mudali (University of Zululand) , Nathi Zulu
(University of Zululand) and Matthew Adigun (University of Zululand) Page 247
Title: 45: Bandwidth Reduction Using Correlated Source Compression for Smart Grid Meters with
Feedback
Authors: Reevana Balmahoon (University of the Witwatersrand) and Ling Cheng (University of the
Witwatersrand) Page 253
Title: 73: Energy Minimization in WSNs: Empirical Study of Multicast Incremental Power Algorithm
Authors: Adeyemi Abel Ajibesin (University of Cape Town), Neco Ventura (University of Cape Town), Alexandru
Murgu (University of Cape Town) and H. Anthony Chan (Huawei Technologies) Page 259
Title: 48: On Rayleigh Approximation of the Multipath PLC Channel: Broadband through the PLC
Channel
Authors: Abraham Nyete (University of KwaZulu-Natal), Thomas J.O. Afullo (University of KwaZulu-Natal) and
Innocent Davidson (University of KwaZulu-Natal) Page 265
Title: 79: Development of an improved routing metric based on IBETX Metric for Wireless Ad-hoc
Networks
Authors: Maxime Kabiwa (Tshwane University of Technology), Karim Djouani (University of Paris-East Creteil)
and Anish Kurien (Tshwane University of Technology) Page 271
Title: 7: Enhanced Backoff Mechanism for the Traditional Carrier Sense Multiple Access with
Collision Avoidance in a IEEE 802.11p VANET
Authors: Ifer Barbana Kam (Tshwane University of Technology), Karim Djouani (Tshwane University of
Technology) and Anish Kurien (Tshwane University of Technology) Page 277
Title: 5 75: Capacity Performance Analysis in MIMO Vehicular Networks
Authors: Ferdinand Nyongesa (Tshwane University of Technology), Karim Djouani (Tshwane University of
Technology) and Alex Hamam (Tshwane University of Technology) Page 283
Title: 5 17: Stock Position Tracking and Theft Prevention System
Authors: Solomon Petrus Le Roux (Stellenbosch University) and Riaan Wolhuter (Stellenbosch University) Page 289
Title: 119: A Link Quality Aware Rumor Based Protocol for Wireless Sensor Networks
Authors: N'Guettia William Kouassi (Tshwane University of Technology), Karim Djouani (Tshwane University of
Technology) and Anish Kurien (Tshwane University of Technology) Page 295
Title: 5 16: Collaborative Incentive Schemes and Virtual Coordinate Routing in Sensor Networks
Authors: Anthony Krzesinski (Stellenbosch University) and Dirk Brand (Stellenbosch University) Page 301
Title: 5 118: Rural Wireless Mesh Network Analysis On-The-Go
Authors: Ghislaine Livie Ngangom Tiemeni (University of the Western Cape), Isabella Venter (University of the
Western Cape) and William Tucker (University of the Western Cape) Page 307
Title: 5 124: Dispersive Characteristics for Broadband Indoor Power-Line Communication Channels
Authors: Modisa Mosalaosi (University of KwaZulu-Natal) and Thomas Afullo (University of KwaZulu-Natal) Page 313
Title: 9: Performance Analysis of Dynamic Switching between Spatial Multiplexing and Diversity over
Rayleigh Fading Channels in MIMO-OFDM Systems using QPSK Modulation SchemeModulation Scheme
Authors: Jamal Ramadan Elbergali (College of Industrial Technology) and Neco Ventura (University of Cape
Town) Page 319
Title: 57: Performance Evaluation of RSSI based CCA-Map Localisation Algorithm in Wireless Sensor Networks
Authors: Omotayo Ganiyu Adewumi (Tshwane University of Technology) Page 325
Title: 43: An investigation into the Accuracy of the Kriging Method for Multiple Wi-Fi Access Point
RSSI Estimation
Authors: PJ Joubert (North-West University) and ASJ Helberg (North-West University) Page 331
Title: 11: Effect of Node Pause Time and Speed on Routing Protocols in Mobile Ad-Hoc Networks
Authors: Telma Nokane Botshelo (North-West University), Michel Mbougni (Concordia University), Obeten
Ekabua (North-West University) and William Montshosi (North-West University) Page 337
6. Management
Title: 44: Congestion Control in Multi-Serviced Heterogeneous Wireless Networks Using Dynamic
Pricing (with Users' Willingness to Pay Incorporation)
Authors: Samson Oluwashina Orimolade (University of Cape Town) and Olabisi Falowo (University of Cape
Town) Page 345
Title: 38: Freight Tracking Cost Analysis to Improve Logistics Management Operations
Authors: Arno de Coning (North-West University) and Alwyn J Hoffman (North-West University) Page 351
Title: 76: CoBI: A Collective Biosignal-Based Identification Model
Authors: Dustin van der Haar (University of Johannesburg) and Sebastiaan von Solms (University of
Johannesburg) Page 357
Title: 40: Perishable Produce Temperature Profiling Using Intelligent Telematics
Authors: Christian Chuks Emenike (North-West University) and Alwyn Jacobus Hoffman (North-West University) Page 363
Title: 5 35: Using Mobile Networks for Effective Cold Chain Management
Authors: Bernardus P. van Eyk (North-West University) and Alwyn J. Hoffman (North-West University) Page 369
Title: 100: On the Optimal Artificial Neural Network Architecture for Forecasting TCP/IP Network
Traffic Trends
Authors: Vusumuzi Moyo (University of Fort Hare) and Khulumani Sibanda (University of Fort Hare) Page 375
7. Standards, Regulatory & Environmental
Title: 1: Prototyping Machine-to-Machine Applications for Emerging Smart Cities in Developing
Countries
Authors: Joyce Mwangama (University of Cape Town), Joseph Orimolade (University of Cape Town), Neco Ventura (University of Cape Town), Asma Elmangoush (Technische Universität Berlin), Ronald Steinke (Technische Universität Berlin), Alexander Willner (Technische Universität Berlin), Andreea Corici (Fraunhofer FOKUS Research Institute) and Thomas Magedanz (Fraunhofer FOKUS Research Institute)
Page 383
Title: 66: Received Power Prediction of a Terrestrial TV Broadcasting Transmitter Using Ordinary
Kriging Interpolation
Authors: Willem Hendrik Boshoff (North-West University), Magdalena Grobler (North-West University) and
Melvin Ferreira (North-West University) Page 389
Title: 84: Standard Compliant Channel Selection Scheme for TV White Space Networks
Authors: Moshe Timothy Masonta (Council for Scientific and Industrial Research), Thomas Olwal (Council for Scientific and Industrial Research), Fisseha Mekuria (Council for Scientific and Industrial Research) and Mjumo Mzyece (Tshwane University of Technology)
Page 395
Title: 19: Improving Trustworthiness amongst Nodes In Cognitive Radio Networks
Authors: Efe Orumwense (University of KwaZulu-Natal), Olutayo Oyerinde (University of the Witwatersrand) and
Stanley Mneney (University of KwaZulu-Natal) Page 401
Page xxiii
Southern Africa Telecommunication Networks and Applications Conference (SATNAC) 2014Title: 83: Analysis of Spectral Opportunity in the UHF Terrestrial TV Frequency Band
Authors: Melvin Ferreira (North-West University) and Albert Helberg (North-West University) Page 407
Title: 24: The Impact of Regulation and Policy on Secondary User Pricing Strategies in a Cognitive
Radio Environment in South Africa
Authors: Elicia Naidu (University of the Witwatersrand) and Rex Van Olst (University of the Witwatersrand) Page 413
Title: 12: Practical Glycerol Water Solution Measurements to Determine the Effects which the Fluid
Properties has on the Drop Formulation Process for 3D Printers
Authors: PJM van Tonder (Vaal University of Technology), HCvZ Pienaar (Vaal University of Technology) and
DJ de Beer (North-West University) Page 419
Title: 5 33: Experimental Assessment of PV Module Cooling Strategies
Authors: Augustine Ozemoya (Vaal University of Technology), James Swart (Vaal University of Technology)
and Christo Pienaar (Vaal University of Technology) Page 425
Title: 14: Quantifying the Effect of Varying Percentages of Full Shading on the Output Power of a PV
Module in a Controlled Environment
Authors: Arthur J Swart (Central University of Technology) and Pierre E Hertzog (Central University of
Technology) Page 431
Work In Progress 1: Access Networks Technologies
Title: 172: Incremental FTTH deployment planning Page 439
Authors: Jonabelle Laureles (North-West University), Leenta Grobler (North-West University) and Fanie Terblanche (North-West University)
Title: 185: An efficient Sum-product decoding algorithm for Quasi-Cyclic LDPC codes Page 441
Authors: Yuval Genga (University of the Witwatersrand)
Title: 139: Optimal QoS Aware scheduling algorithm for improved inter cell interference in LTE
system Page 443
Authors: Evah Mmatsatsi Mthimunye (Tshwane University of Technology) and Prof. Karim Djouani (Tshwane University of Technology) and Prof. Anish Kurien (Tshwane University of Technology)
Work In Progress 2: Converged Services
Title: 150: A Single-Queue Priority Scheduler for Video Transmission in WLANs Page 447
Authors: Joshua Adeleke (University of Cape Town), Mqhele Dlodlo (University of Cape Town) and Clement Onime (ICTP)
Title: 175: A Connection Management System to Enable the Wireless Transmission of MIDI Messages Page 449
Authors: Brent Shaw (Rhodes University) and Richard Foss (Rhodes University)
Title: 132: Real-time Background Subtraction Under Sudden Illumination Changes Page 451
Authors: Cornelius JF Reyneke (University of Johannesburg), Philip E Robinson (University of Johannesburg) and André L Nel (University of Johannesburg)
Title: 167: Autonomous Facial Expression Recognition Using the Facial Action Coding System Page 453
Authors: Nathan de la Cruz (University of the Western Cape), Mehrdad Ghaziasgar (University of the Western Cape), James Connan (University of the Western Cape) and Reg Dodds (University of the Western Cape)
Title: 134: CUDA Optimized Dynamic Programming Search for Automatic Speech Recognition on a
GPU Platform Page 455
Authors: Babedi Betty Letswamotse (North West University), Naison Gasela (North West University) and Zenzo Polite Ncube (Sol Plaatjie University)
Work In Progress 3: Core Network Technologies
Title: 163: Towards Investigating Transmission Penalties in a Flexible Spectrum Optical Network Page 459
Authors: Duncan Kiboi Boiyo (Nelson Mandela Metropolitan University), Enoch Rotich Kipnoo (Nelson Mandela Metropolitan University), Romeo Gamatham (Nelson Mandela Metropolitan University), Andrew Leitch(Nelson Mandela Metropolitan University) and Tim Gibbon (Nelson Mandela Metropolitan University)
Work In Progress 4: Data Centre and Cloud
Title: 149: Micro Data Centres for Multi-Service Access Nodes and their effect on latency and
services Page 463
Authors: David van Wyk (University of Pretoria) and Jacques van Wyk (University of Pretoria)
Work In Progress 5: Internet Services & End User Applications
Title: 164: Using a Natural User Interface to Support Information Sharing Among Co-Located Mobile
Devices Page 467
Authors: Timothy Lee Son (Nelson Mandela Metropolitan University), Janet Wesson (Nelson Mandela Metropolitan University) and Dieter Vogts (Nelson Mandela Metropolitan University)
Title: 131: Preliminary Thoughts on Services without Servers Page 469
Authors: Philip Machanick (Rhodes University) and Kieran Hunt (Rhodes University)
Title: 184: Efficiency of Mobility Management schemes on virtualized shared Mobile Networks Page 471
Authors: Ofentse Noah (University of Cape Town) and Neco Ventura (University of Cape Town)
Title: 144: A Model of Fraud Detection in Mobile Transaction via Unstructured Supplementary Service
Data Page 473
Authors: Kulani Vukeya (Tshwane University of Technology) and Paul O. Kogeda (Tshwane University of Technology)
Work In Progress 6: Limited Range Communications
Title: 143: An arrival-time Detection Technique for Multilateration-based Automatic Wildlife Tracking Page 477 Authors: Schalk-Willem Krüger (North-West University) and Albert Helberg (North-West University)
Title: 127: AntMAC: A Dynamic Control Channel Selection MAC Protocol Design for Cognitive Radio
Ad Hoc Network Page 479
Authors: Henry Ohize (University of Cape Town) and Mqhele Dlodlo (University of Cape Town)
Title: 135: The Design of an Energy Efficient Routing Protocol for a Wireless Ad-hoc Mesh Network
for Animal Tracking Page 481
Authors: Charles James van der Spuy (North-West University) and Albertus Stephanus Jacobus Helberg (North-West University)
Title: 178:A Comparison of Wireless Sensor Network Routing Protocols on a Low Cost Prototype
Experimental Test Bed Page 483
Authors: John George Moutzouris (University of the Witwatersrand)
Work In Progress 7: Management
Title: 177:Application-based Network Policy Management for Machine-to-Machine Devices Page 487 Authors: Nyasha Mukudu (University of Cape Town) and Neco Ventura (University of Cape Town)
Work In Progress 8: Standards, Regulatory & Environmental
Title: 182: Analysis and Design of Wireless Mobile Charging System for Mobile Phones Using
Communication Networks Page 491
Authors: Frederick Gyampoh Kumi (University of Cape Town) and Mqhele E. Dlodlo (University of Cape Town)
Page xxv
Southern Africa Telecommunication Networks and Applications Conference (SATNAC) 2014Page xxvii
Southern Africa Telecommunication Networks and Applications Conference (SATNAC) 2014Optimal Passive Optical Network Planning Under
Demand Uncertainty
S.P. van Loggerenberg
†, M. Ferreira
†, M.J. Grobler
†, S.E. Terblanche
‡TeleNet Research Group
†
School of Electrical, Electronic and Computer Engineering
‡Centre for Business Mathematics and Informatics
North-West University, Potchefstroom Campus, South Africa
Email: {20289278, melvin.ferreira, leenta.grobler, fanie.terblanche}@nwu.ac.za
Abstract—As a result of ever-increasing demand for access level bandwidth, long deployment cycles and the popularisation of more economically viable Point-to-Multipoint (P2MP) networks, service providers are moving to extensively future-proof fibre technologies to connect consumers. Of these, the Passive Optical Network (PON) is the most prevalent. Though the optimal planning of these networks have been studied by a number of authors recently, the typical situation where consumer demand is uncertain has yet to be addressed. By including stochastic demand in an Integer Linear Program (ILP) model through the use of 2-stage stochastic programming, this can be accounted for. In this paper, a discrete approach is followed, optimising the model with the addition of consumer demand scenarios using real-world Geographic Information System (GIS) data. Results show a definitive decrease in deployment cost when any of the scenarios realise, especially when splitter capacity is restrictive.
Index Terms—ILP, Network Modelling, Optimisation, PON, Stochastic Modelling
I. INTRODUCTION
A
CCORDING to [1], international bandwidth demand grew by 39 % during 2012, with Africa’s bandwidth demand projected to grow annually at a rate of 51 % [2] between 2012 and 2019. This increase in demand is mostly attributed to the rise in popularity of access-level streaming video and requires extensively future-proof technologies, due in part to inability of service providers to roll out new technologies every year. Though access network technologies exist to fill the gap between current demand and short-term projected demand, such as Very-high-bit-rate Digital Subscriber Line (VDSL) and Long Term Evolution (LTE), they do not have the required bandwidth potential to keep up with long-term consumer bandwidth requirements. Due to recent standard advancements in fibre technology, P2MP fibre networks such as PON have become an economically viable alternative to copper, providing high bandwidth, noise-immune access networks from street level (Fibre to the Curb (FTTC)), right up to customer premises (Fibre to the Home (FTTH)).The two main standards for PON are ITU-T G.984 Gigabit Passive Optical Network (GPON) [3] and IEEE 802.3ah Eth-ernet Passive Optical Network (EPON) [4]. EPON builds on the existing Ethernet standard, allowing quick integration with existing network infrastructure while providing bidirectional bandwidth of 1 Gb/s. GPON is based on Asynchronous
Transfer Mode (ATM), providing legacy support while increas-ing up- and downstream bandwidth rates to 1.244 Gb/s and 2.448 Gb/s, respectively. Recently, 10 Gb/s versions of both standards have also been ratified, with G.987 XG-PON [5] and 802.3av 10G-EPON [6].
Planning of PON networks remains a challenge, with sub-optimal manual plans leading to unnecessarily inflated deploy-ment costs, potentially in the order of millions of rands. Even though a number of approaches have been followed to pro-duce optimal or close-to-optimal plans for P2MP fibre access networks [7]–[10] (see [11] for a more in-depth summary), even including fibre duct sharing [12], there still exists a gap between these approaches and real-world deployments: demand uncertainty. When service providers plan greenfield fibre networks, i.e. trenching and laying fibre where none ex-isted before, the demand of consumers in the area is unknown, introducing a large uncertainty into the eventual utilisation of the network and effectively the Return on Investment (ROI). In this paper we will incorporate demand uncertainty into the PON planning model itself to improve solution quality over a number of possible outcomes, which has, to the best of our knowledge, not been done before.
The rest of the paper is organised as follows: Section II de-fines the PON planning problem in greater detail with respect to the topology and assumptions made. Section III introduces the concept of demand uncertainty and how it is handled in the planning model with section IV defining the resulting mathematical model. In section V, the uncertainty model is compared to a standard planning model before concluding the paper in section VI with a summary and suggestions for future work.
II. PROBLEMDEFINITION A. Topology
PON follows a hierarchical tree topology with the Central Office (CO) as the root node. Fibres run from an Optical Line Terminal (OLT) at a CO to a number of splitters, where the optic signal is passively split into a number of downstream signals. These passive splitters are then each connected by fibre to a number of Optical Network Units (ONUs) at customer premises. Figure 1 illustrates the PON topology. In
Feeder fiber Distribution fiber Central office ONU Splitter Fiber-to-the-Building (FTTB) Fiber-to-the-Home (FTTH)
Fig. 1: Basic Passive Optical Network topology
this paper, the network of fibres between COs and splitters is called the feeder network, whereas the fibres between splitters and ONUs are collectively known as the distribution network. The greatest advantage of PON is that a single feeder fibre from the CO serves a number of ONUs down the line, greatly reducing overall network cost. The GPON standard specifies the maximum number of ONUs that can be served from a single fibre, also known as the maximum split ratio, as 1:128, while EPON supports as many as the link budget allows, usually up to 1:64 [13]. These high split ratios require expensive high power optics to attain a feasible network reach, therefore split ratios of 1:16 and 1:32 are more common.
Given a number of ONU locations at customer premises, a CO location and a number of potential sites for splitters, the PON planning problem then becomes the search for a network that simultaneously solves the following [14]:
• Determining the optimal number of splitters, • Allocating ONUs to splitters,
• Relocating splitters,
• Routing and aggregating fibre to minimize overall de-ployment cost.
Unfortunately, the structure of the problem makes solving this model extremely hard, known as Non-deterministic Poly-nomial time (NP)-hard in complexity theory. In other words, it has been mathematically proven that it is impossible to construct a deterministic algorithm that can solve this problem to optimality in polynomial time, or O(nk), with a fixed k [14]. This behaviour ensures that anything but the smallest of datasets can not be solved in a feasible time.
B. Model assumptions
Since the problem is so difficult to solve, a number of assumptions have been made to reduce complexity to a manageable level. Firstly, a few network constraints will be ignored for the model in question, including the placement of splicing boxes as well as the actual power budget constraints. Therefore, for this proof of concept model, it is assumed that individual lossless fibres can be placed in a duct without the need for splicing. The deployment area is also assumed to be greenfield, with no existing ducts available.
To simplify costing, a fixed cost per unit length of trenching and fibre is assumed, as well as a single splitter type. Further-more, since we want to mainly test the distribution network side as this is where the uncertainty lies, we assume a single CO is available.
It should be noted that even though these factors are omitted in this paper, they can in fact be modelled (as was shown in our work in [15]).
C. Input data
As input for PON planning model, most authors use ran-domly generated datasets with fixed distributions. This pro-vides decent performance indication, but can hide potential weaknesses in the planning approach. To avoid this potential bias, we use real-world GIS data of a typical PON network topology. These datasets contain a set of geo-referenced points or nodes, defined by latitude and longitude, each connected to subset of neighbouring points with a set of edges, each with its own weight (in our case, the distance between the nodes). To determine the distance between two non-connected nodes, a shortest-path algorithm like Dijkstra’s algorithm can be used. In [16], the author provides an algorithm based on Dijkstra that outputs the k shortest paths between two nodes on a graph. This algorithm can also give all the possible paths between two nodes if it terminates before reaching an arbitrarily large value for k, which is the configuration required for the provided models. Note that a heuristic solution can be computed by including only a subset of all the possible paths [12].
III. DEMANDUNCERTAINTY
As we mentioned before, for greenfield network deploy-ment, it is quite rare that consumer demand is known at the planning stage. Even though the network could be designed based on the notion that every consumer will use the PON service, it is unrealistic and can lead to unnecessary expenses during deployment. Therefore, consumer demand has to be estimated or projected based on certain known metrics, e.g. demand for previous services or average income per house-hold, to be able to plan a network accordingly.
To model demand uncertainty, we utilise a widely used technique known as 2-stage stochastic programming (see [17, p. 103]). This technique uses data available at the moment the decision is made (first stage) along with some recourse action that estimates the effect when the uncertainty is revealed (second stage). In this paper, we consider a discrete approach based on independent scenarios. In this approach, a scenario is a possible outcome of the consumer demand, containing a subset of the total demand with some given probability. Figure 2 shows two scenarios as an example.
A vital assumption for this approach is that only one scenario realises at the end, allowing the model to over utilise splitter capacity between scenarios for lower deployment cost. Therefore the probabilities of all scenarios realizing should sum to 100 %. For simplicity, in this paper we will assume each scenario has an equal probability to realise.
A B E Scenario 2 A B C F Scenario 1 A B C D E F Total demand
Fig. 2: Example of two scenarios
For the PON planning approach contained in this paper, we assume that the planning occurs in two phases. In the first phase, it is assumed that the service provider uses a plan to determine where to trench for fibre ducts. In the second phase, a demand scenario realises and the service provider proceeds to place fibre in the existing trenches where needed, effectively working with a brownfield network. Note that this second phase should not be confused with the recourse stage of the 2-stage stochastic programming approach, as they function externally and internally to the model respectively.
Unfortunately, this approach inevitably leads to increased complexity as we have to effectively enumerate a large number of possibilities for the uncertain demand to realise. This aspect of the problem, scenario generation, is not discussed in detail in this paper, although a number of techniques exist to generate independent scenarios, including using historical data or Monte Carlo methods [18], [19].
IV. MATHEMATICALMODELS
Three mathematical models are necessary to illustrate the workings and advantages of a 2-stage stochastic model. We will now discuss each of these below, by first introducing the required notation and then detailing the model formulation.
A. Sets
• U - A set of all possible locations for ONUs in the form of Cartesian coordinates. The index j is used to indicate the coordinates of the j-th ONU.
• D - The set of all possible locations for splitters in the model. For this set, the index i is used to indicate the i-th splitter coordinates.
• S - The set of all scenarios. Index s is used for the s-th scenario.
• K - The set of all commodity pairs, i.e. all combinations of the CO and splitters and splitters and ONUs.
• P - The set of all possible paths between all commodity pairs.
• E - The set of all edges contained in the input graph.
B. Subsets
• KF(i) ⊆ KF⊂ K - Subset of all feeder commodity pairs between the CO and a splitter that contains the element i, i.e. all combinations of CO and splitters that contain either CO i or splitter i.
• KD(i) ⊆ KD⊂ K - Subset of all distribution commod-ity pairs between a splitter and an ONU that contains the element i, i.e. all combinations of splitters and ONUs that contain either splitter i or ONU i. For example KD(j), j ∈ U, contains all combinations of splitters and the j-th ONU.
• PD(k) ⊆ P - A subset containing all possible paths between the distribution commodity pair k ∈ KD. • PF(k) ⊆ P - A subset containing all possible paths
between the feeder commodity pair k ∈ KF.
• P(e) ⊆ P - A subset consisting of all paths that contain the edge e ∈ E.
• P(k, e) ⊆ P(e) - A subset consisting of all paths between commodity pair k ∈ K that contain the edge e ∈ E. C. Variables
• ypsD - Binary variable used to indicate usage of the p-th distribution pap-th for scenario s, p ∈ P, s ∈ S. The variable takes on a value of 1 if the p-th path is used for scenario s and 0 if it’s unused.
• ypF - Binary variable used to indicate usage of the p-th feeder path, p ∈ P. The variable takes on a value of 1 if the p-th path is used and 0 otherwise.
• yˆpD - Binary variable used to indicate usage of the p-th distribution pap-th across all scenarios, p ∈ P. The variable takes on a value of 1 if the p-th path is used and 0 otherwise.
• yp - Binary variable used to indicate usage of the p-th path, p ∈ P. The variable takes on a value of 1 if the p-th path is used and 0 otherwise.
• xe - Binary variable used to indicate usage of the e-th edge, e ∈ E. Similarly, the variable takes on a value of 1 if the e-th edge is used and 0 if it’s unused.
• ψi - Binary variable used to define the usage of the i-th splitter, i ∈ D. If the splitter is used, the variable takes on a value of 1. If unused, the variable is 0.
D. Parameters
• cCO - The fixed OLT cost incurred for each PON at the CO.
• cSP- The cost associated with deploying a single splitter. • cONU - The cost to deploy a single ONU in the field. • cT - Average cost per meter of trenching.
• cD - Average cost per meter of distribution fibre. • cF - Average cost per meter of feeder fibre.
• σjs - Takes on the value of 1 if ONU j is contained in scenario s ∈ S and 0 otherwise, j ∈ U.
• κ - The maximum number of ONUs that can connect to a single splitter, i.e. splitter capacity or maximum split ratio.
• `e, `p- The length in meter of the e-th edge and p-th path respectively, e ∈ E, p ∈ P.
E. Model Formulation
1) Deterministic model (DETRM): Firstly, we present an ILP model for the PON planning problem without taking uncertain demand into account, i.e. a deterministic model (henceforth DETRM). This model is a standard formulation as seen in a number of works [14], [20], with the addition of fiber duct sharing as shown in [12].
min cCO+ X i∈D ψicSP+ |U|cONU+ X p∈PD yp`pcD + X p∈PF yp`pcF+ X e∈E xe`ecT (1) s.t. X k∈KD(j) X p∈PD(k) yp= 1 ∀j ∈ U (2) X k∈KF(i) X p∈PF(k) yp= ψi ∀i ∈ D (3) X k∈K X p∈P(k,e) yp≤ |P(e)|xe ∀e ∈ E (4) X k∈KD(i) X p∈PD(k) yp≤ κψi ∀i ∈ D (5)
The model minimises total deployment cost in (1), con-sisting of the fixed OLT cost, splitter deployment cost, ONU cost, distribution and feeder fibre cost, and trenching cost. Constraints (2) and (5) specify the distribution side of the PON network, including the fibre demand for each ONU and the maximum number of ONUs that can connect to each splitter. Equation (5) also ensures that the splitter usage variable ψi is set if one or more ONUs are connected to a splitter i. Constraint (3) ensures that a fibre is present between each used splitter and the CO, while the inequality in (4) ensures that a trench is placed if at least one fibre crosses the edge.
2) 2-stage stochastic model (2STOCH): By including un-certain demand through the use of the 2-stage stochastic programming approach, we arrive at the following ILP model (henceforth 2STOCH). min cCO+ X i∈D ψicSP+ |U|cONU+ X p∈PD ˆ ypD`pcD + X p∈PF yFp`pcF+ X e∈E xe`ecT (6) s.t. X k∈KD(j) X p∈PD(k) ypsD = σjs ∀j ∈ U, ∀s ∈ S (7) X k∈KF(i) X p∈PF(k) yFp = ψi ∀i ∈ D (8) X k∈KD X p∈PD(k,e) ˆ yDp + X k∈KF X p∈PF(k,e) yFp ≤ |P(e)|xe ∀e ∈ E (9) X s∈S ypsD ≤ |S|ˆyDp ∀p ∈ PD(k), ∀k ∈ KD (10) X k∈K (i) X p∈P (k) yDps≤ κψi ∀i ∈ D, ∀s ∈ S (11)
The objective function (6) of the model is the same as that of DETRM, since we also want to minimise the total deployment cost. Constraint (7) is adapted from (2) but now specifies the demand for ONU j in scenario s. Equation (8) is identical to (3), while (9) is adapted to include a duct when a fibre from any scenario passes through an edge. The inequality (10) ensures that when a specific path is used in any scenario, only a single fibre is included in the total plan. This is valid since only a single scenario can realise eventually. Finally, (11) ensures a splitter is used when an ONU is allocated to it from within any scenario.
3) Realisation model (REAL): If we were to deploy the complete networks as given in the solutions to DETRM and 2STOCH, the network would likely be impractically oversized. This is due to excess capacity reserved for ONUs not contained in the realised scenario. Therefore, to test the practical cost of the plan, we need to be able to determine the minimum cost of the network, given a scenario s. This solution represents the second phase of deployment, using ducts that have been de-ployed according to the solutions of DETRM and 2STOCH in the first phase. At this point, both splitter and fibre placement need to be determined to complete the network. This model will henceforth be called the realisation model (REAL).
¯
X donates a subset of the set X, e.g. ¯Us is the subset of U containing only the ONUs from scenario s ∈ S and ¯PD is the set of distribution paths between splitters and ONUs contained in scenario s. Similarly, the subset ¯K contains all commodities present in scenario s and ¯E contains all edges used in paths between commodities in ¯K.
min cCO+ X i∈D ψicSP+ | ¯Us|cONU+ X p∈ ¯PD yp`pcD + X p∈ ¯PF yp`pcF+ X e∈ ¯E xe`ecT (12) s.t. X k∈ ¯KD(j) X p∈ ¯PD(k) yp= 1 ∀j ∈ ¯Us (13) X k∈ ¯KF(i) X p∈ ¯PF(k) yp= ψi ∀i ∈ D (14) X k∈ ¯K X p∈ ¯P(k,e) yp ≤ | ¯P(e)|xe ∀e ∈ ¯E (15) X k∈ ¯KD(i) X p∈ ¯PD(k) yp≤ κψi ∀i ∈ D (16)
REAL is identical in structure to DETRM, but uses subsets depending on the scenario that realises. This model therefore effectively solves a complete brownfield network.
V. RESULTS ANDANALYSIS
All models were implemented in C++ using the Concert extensions of IBM ILOG CPLEX and solved on an Intel Core i7 @ 2.67 GHz with 16 GiB memory running Windows.
Demand uncertainty is tested by solving both DETRM and 2STOCH to optimality. Then, REAL is solved for each scenario in S, using the ducts from DETRM and 2STOCH.
To ensure the models can be solved in a feasible time, a very small GIS-mapped dataset known as MicroNet is used,
TABLE I: Total deployment cost results with κ = 8
Model Scenario Objective (R) Splitters used
DETRM 178,301.52 2 DETRM + REAL 1 100,456.66 1 DETRM + REAL 2 146,382.39 1 DETRM + REAL 3 72,152.84 1 2STOCH 171,851.24 1 2STOCH + REAL 1 100,456.66 1 2STOCH + REAL 2 146,382.39 1 2STOCH + REAL 3 72,152.84 1
TABLE II: Total deployment cost results with κ = 4
Model Scenario Objective (R) Splitters used
DETRM 202,565.98 3 DETRM + REAL 1 106,935.77 2 DETRM + REAL 2 159,882.78 2 DETRM + REAL 3 72,152.84 1 2STOCH 179,101.68 2 2STOCH + REAL 1 106,906.94 2 2STOCH + REAL 2 154,433.00 2 2STOCH + REAL 3 72,152.84 1
containing 10 ONUs, 3 possible splitter locations and 36 edges. Three scenarios were manually generated, containing between 3 and 6 ONUs each. To allow a fair comparison in deployment cost between DETRM and 2STOCH, it was ensured that every ONU is contained in at least one scenario. Maximum split ratio was set to either 1:8 or 1:4 to test its influence on deployment cost. Even though such low maximum split ratios are rarely seen in practical network sizes, it is necessary to scale the constraint to the dataset size to illustrate the effect. In this case, if a split ratio of 1:16 were allowed, all ONUs could be connected to a single splitter, which will almost never be the case in practice.
The total deployment cost of each test run is given in tables I and II. Given a maximum split ratio of 1:8, DETRM gives a 3.8 % higher objective value than 2STOCH, even though all scenarios give the same deployment cost for both types of models. The ducts used in the solutions to both DETRM and 2STOCH are the same, even though fibres are connected differently, explaining the identical scenario results. DETRM also uses an extra splitter, indicating that extra capacity is reserved in the model, even though it is never used when a single scenario realises.
When the split ratio is more tightly constrained to a max-imum of 1:4, the difference between models become more evident, with the total deployment cost of DETRM now 13.1 % higher than 2STOCH. Since the resulting network plan for DETRM and 2STOCH differs quite dramatically at this split ratio, the ducts available to REAL changes as well. This results in a slight increase in deployment cost of 0.02 % for scenario 1 and a substantial 3.5 % for scenario 2 when demand uncertainty is excluded, even though the same number of splitters are used.
Figures 3 and 4 show the resulting PON topology for DETRM and 2STOCH respectively, using a maximum split ratio of 1:4. In these figures, circles represent ONUs, triangles represent used splitters and the square represents the CO. Used
Fig. 3: Optimal solution for DETRM with κ = 4
Fig. 4: Optimal solution for 2STOCH with κ = 4
trenches are coloured green through yellow to red, indicating, in ascending order, the number of fibres contained in each. Grey components indicate unused ducts or splitters. From the figures it is clear that DETRM tries to connect all ONUs without regarding the independent nature of scenarios.
VI. CONCLUSION& FUTUREWORK
In this paper, demand uncertainty was incorporated in a model of the PON planning problem through the use of a discrete 2-stage stochastic programming approach. This ensures that a finite number of scenarios can be calculated to predict consumer demand, allowing for splitter over-utilisation and optimal duct planning. To the best of our knowledge, this type of formulation has not been previously published, likely due to the increased complexity that is inevitable when modelling stochastic parameters.
Using a small dataset, it was evident that the gains from this approach are very much dependent on the splitter capacity, with tighter restrictions allowing for more savings, up to 13 %. When re-optimising the second phase deployment, gains of up to 4 % were demonstrated. These figures are heavily dependent on the data, with larger datasets expected to show much larger savings due to the increased number of possible fibre paths.
The advantage of this approach stems from the fact that scenarios are independent, ensuring that excess capacity on
splitters are kept to a minimum and ducts are shared as much as possible across all scenarios.
Future work will include algorithmic and decomposition methods to improve computational performance of the mod-elling approach provided, allowing for much larger datasets to be solved. Furthermore, though scenario generation is discussed in detail in other works, it would be of interest to determine how the process can be tailored for the PON planning problem in particular. Handling of unequal scenario probabilities is also necessary to improve real-world applica-bility. Finally, a number of refinements can be made to the model to remove some of the more restrictive assumptions, including the use of multiple splitter types, multiple COs and network constraints such as optical power budget.
ACKNOWLEDGEMENT
The authors thank the anonymous reviewers for their con-structive feedback and acknowledge the financial support of the Telkom Centre of Excellence (CoE) at the North-West University, Potchefstroom Campus.
REFERENCES
[1] TeleGeography. (2013, April) International
band-width demand is decentralizing. [Online].
Available: http://www.telegeography.com/press/press-
releases/2013/04/17/international-bandwidth-demand-is-decentralizing/index.html
[2] TeleGeography. (2013, October) Africa’s international bandwidth demand to lead the world. [Online]. Available: http://www.telegeography.com/press/press-releases/2013/10/31/africas-international-bandwidth-demand-to-lead-the-world/index.html [3] ITU-T, Gigabit-capable passive optical networks (GPON): General
characteristics, Mar 2008, ITU-T Rec. G.984.1. [Online]. Available: http://www.itu.int/rec/T-REC-G.984.1/en
[4] IEEE, Standard for Information Technology- Telecommunications and Information Exchange Between Systems- Local and Metropolitan Area Networks- Specific Requirements Part 3: Carrier Sense Multiple Access With Collision Detection (CSMA/CD) Access Method and Physical Layer Specifications Amendment: Media Access Control Parameters, Physical Layers, and Management Parameters for Subscriber Access Networks, IEEE Std. 802.3ah, 2004.
[5] ITU-T, 10-Gigabit-capable passive optical networks (XG-PON): General requirements, Jan 2010, ITU-T Rec. G.987.1. [Online]. Available: http://www.itu.int/rec/T-REC-G.987.1/en
[6] IEEE, Standard for Information technology - Telecommunications and information exchange between systems - Local and metropolitan area networks - Specific requirements Part 3: Carrier Sense Multiple Access with Collision Detection (CSMA/CD) Access Method and Physical Layer Specifications Amendment 1: Physical Layer Specifications and Management Parameters for 10 Gb/s Passive Optical Networks, IEEE Std. 802.3av, 2009.
[7] M. Gr¨otschel, C. Raack, and A. Werner, “Towards optimizing the deployment of optical access networks,” ZIB, Takustr.7, 14195 Berlin, Tech. Rep. 13-11, 2013.
[8] M. Leitner, I. Ljubic, M. Sinnl, and A. Werner, “On the two-architecture connected facility location problem,” ZIB, Takustr.7, 14195 Berlin, Tech. Rep. 13-29, 2013.
[9] I. Ljubi´c and S. Gollowitzer, “Layered graph approaches to the hop constrained connected facility location problem,” INFORMS J. on Com-puting, vol. 25, no. 2, pp. 256–270, Apr. 2013.
[10] M. Chardy, M.-C. Costa, A. Faye, and M. Trampont, “Optimizing splitter and fiber location in a multilevel optical FTTH network,” European Journal of Operational Research, vol. 222, no. 3, pp. 430 – 440, 2012. [11] S. van Loggerenberg, M. Grobler, and S. Terblanche, “Optimization of pon planning for ftth deployment based on coverage,” in South-ern African Telecommunications and Networks Access Conference (SATNAC), 2012 Proceedings of, Sep. 2012.
[12] S. van Loggerenberg, M. Grobler, and S. Terblanche, “Optimization of pon planning for ftth deployment with fiber duct sharing,” in Southern African Telecommunications and Networks Access Conference (SATNAC), 2013 Proceedings of, Sep. 2013.
[13] F. Effenberger, D. Clearly, O. Haran, G. Kramer, R. D. Li, M. Oron, and T. Pfeiffer, “An introduction to pon technologies [topics in optical communications],” Communications Magazine, IEEE, vol. 45, no. 3, pp. S17 –S25, March 2007.
[14] J. Li and G. Shen, “Cost minimization planning for greenfield passive optical networks,” Optical Communications and Networking, IEEE/OSA Journal of, vol. 1, no. 1, pp. 17 –29, June 2009.
[15] S. van Loggerenberg, “Optimization of passive optical network planning for fiber-to-the-home applications,” Master’s thesis, North West Univer-sity Potchefstroom Campus, April 2013.
[16] J. Y. Yen, “Finding the k shortest loopless paths in a network,” Management Science, vol. 17, no. 11, pp. pp. 712–716, 1971. [17] J. R. Birge and F. Louveaux, Introduction to stochastic programming.
Springer, 2011.
[18] W. R¨omisch, “Scenario generation,” Wiley Encyclopedia of Operations Research and Management Science, 2011.
[19] M. Kaut and S. W. Wallace, “Evaluation of scenario-generation methods for stochastic programming,” Stochastic Programming E-Print Series, vol. 14, 2003. [Online]. Available: http://edoc.hu-berlin.de/browsing/speps/
[20] K. Poon, D. Mortimore, and J. Mellis, “Designing optimal ftth and pon networks using new automatic methods,” in Access Technologies, 2006. The 2nd Institution of Engineering and Technology International Conference on, June 2006, pp. 49 – 52.
Samuel van Loggerenberg is a Telkom CoE student, currently pursuing his Ph.D in Computer and Electronic Engineering at the North-West University. He received his M.Eng (cum laude) in 2013, B.Eng in 2011 and a B.Sc in Business Mathematics and Informatics in 2008 from the same institution. His research interests include network optimisation, optical networks and peer to multi-peer video streaming networks.
