Addressing the Digital Divide: LR-PON Planning for Sparsely Populated Areas
Post on 17-Jul-2015
103 Views
Preview:
Transcript
Addressing the Digital Divide: LR-PON Planning for Sparsely
Populated Areas
Saptadeep PalCezary ZukowskiAvishek Nag
David B. PayneMarco Ruffini
1
Background and Overview
2
§ Sparse Popula,on : Density of poten,al customers is usually very low Cost of Deployment per user increases
§ User premises are distributed over a large geographical area separated by larger distances.
Fibre length per user is high
§ Generally, these areas are far away from urban centers More fibre required to connect to metro/core nodes
§ Due to lesser take-‐rate and increased infrastructure deployment per user,network operators are reluctant to build FTTH networks for rural areas.
3
Background and Overview
Suggested Approaches towards Rural FTTH Network
4
§ GPON extenders are typically used to extend the reach. But such extenders oIen carry OEO regenerators and thus cost increases.
§ Using Raman Amplifier and SOA at CO
§ Using dual-‐wavelength for upstream and downstream
§ OFDM over WDM/TDM for beRer transmission over longer distances.
§ 1024-‐split architecture for a reach of 100 km (90 km –Backhaul + 10 Km distribu,on sec,on). But in rural areas, ge]ng 1000 customers within a span of 10 Km is oIen very difficult
Courtesy: Telnet
Motivation
5
§ Telecom operators are reluctant to build fibre networks in less dense rural loca,ons.
§ But recently, discussions on deploying the fibre network with the help of exis,ng infrastructure from power grids, rail networks, motorways has come up.
§ LR-‐PON with 3-‐way spli]ng is favoured solu,on for the industry:
Ø Longer range than GPON or Extended-‐GPON.Ø Split ra,o is significantly higher than that of GPON (1024 vs 128).Ø More consolida,on of central offices into huge metro nodes leads to a simpler network and also flexibility for future extension.
§ While 60 – 70 % of the network deployment cost is incurred in trenching while laying the fibre cables.
Ø In this work, we have tried to minimize the fibre cables instead of total fibre length.
ADSL2 Coverage
VDSL Coverage
Rural LR-PON Planning
6
§ LR-‐PON layout is a “lollipop model” that uses a maximum feeder fibre length of 90 km and distribuKon secKon of 10 km .The maximum number of ONUs per PON wavelength is typically up to 1024.
§ But, in sparse rural areas it will be necessary to connect to customers at different points down the feeder route.
§ The Op,cal-‐Distribu,on-‐Network (ODN) reach needs to be extended and alterna,ve configura,ons are considered with longer distribuKon secKon and shorter feeder.
§ In such a case, where the fibre losses in ODN secKon will be more, the number of splits needs to be reduced.
LR-PON Split vs ODN Reach
7
§ The figure shows how the LR-‐PON split ra,o declines with the increase in the ODN reach.
§ In this work, we have used this knowledge to plan the network and maximize resource usage in sparsely populated areas.
§ So, for a strategic deployment in rural areas, a clustering algorithm is required to decide the number of ONUs per spliRer.
Agglomerative Clustering Algorithm
8
§ The algorithm takes the loca,on of the user premises (ONUs) and groups them into capacitated clusters to achieve maximum u,liza,on of the spliRer. The algorithm runs in stages
§ The algorithm first tries to place spliRers with the largest split (32-‐way split). The largest split has the least span. This largest spliRer posi,on will then be the loca,on of the cabinet housing
§ Subsequently, these housing posi,ons will be then used to host other smaller size spliRers to connect the users who could not be reached due to limita,ons in the reach of larger spliRers, thus leading to agglomera,on of more than one type of spliRers at a certain geographical loca,on
§ The algorithm looks to place the housings in the denser areas and build the network around these centers
§ In each itera,on, the algorithm tries to maximize the u,liza,on of each of the spliRers, thus searching for the op,mum loca,on of placing the housings
9
• The Red links are links from the 32-‐way split
• The Yellow ones are links from the 16-‐way spliRer
• The orange ones are from the spliRers with less than 16-‐way spli]ng
• It can be clearly no,ced that most of the cabinet housings with larger spliRers are located in the denser regions.
Minimization of Cable Length
10
We then approach the cable length minimizaKon problem using an ILP and a heurisKc. Cable deployment follows the street layout (taken from the open source open maps database). Close to a user premises, a final drop cable is branched off the public roads to connect the individual user. We call this branching point the final drop point. The link from there to the user premises is normally achieved with an aerial cable.
Representa,on of main roads, ONUs and Delivery points (white circles with black dot on streets)
11
The informa,on about the spliRer posi,on and the ONUs to be served by the spliRer is provided by the clustering algorithm and forwarded as input to the heuris,c. The heuris,c also considers the street mapswhile considering the cable deployment.
Cable Length MinimizaKon HeurisKc
Cable Length Minimization Heuristic
12
Streets in red and ONUs in blue bots
13
The spliRer posi,on is determined by the agglomera,ve clustering algorithm. Firstly, our heuris,c finds the nearest point on a main street for each of the ONUs (i.e., the final drop points) similar to the ILP model and the drop points on same street are joined together.
Cable Length Minimization Heuristic
14
The street segments adjoining the spliRer are joined to the spliRer.
Cable Length Minimization Heuristic
15
Now the connected segments are recursively connected to the other segments which are required to be connected. Note that in this case, one segment might be connected to the more than one already connected segment, we only consider the shortest connec,on.
Cable Length Minimization Heuristic
16
Layout aIer elimina,ng the loops.
Cable Length Minimization Heuristic
17
Final Layout
Cable Length Minimization Heuristic
Test Configuration & Results
18
Major SpliRer Minor SpliRer
Scenario1 S10max =32, R10max =1km S11max = 16, R11max = 12kmScenario2 S10max =32, R10max =2km S11max =16, R11max =11km
Sample Statistics of Cable length Minimization
19
§ Though Dijsktra Algorithm results in about 15% lesser total fibre required, our proposed algorithm significantly decreases the amount of total fibre cable used by about 24% and 30% respec,vely.
§ The proposed heuris,c is approximately 6 – ,mes faster than the ILP while the heuris,c performance as good as that of ILP’s with approximately 5% varia,on in the results.
20
Thank You!
top related