NTu2J.4.pdf OFC/NFOEC Technical Digest © 2013 OSA Cost Optimization in an FTTP Environment Roger L. Tobin Verizon Technology, 60 Sylvan Rd, Waltham, MA 02454 firstname.lastname@example.org Abstract: Verizon has developed several optimization methodologies for the design and operation of the FTTP network. Two of these methodologies, one design and one operations, will be presented and the benefits of advanced analytics discussed. OCIS codes: (060.4256) Network Optimization; (060.1155) All optical networks 1. Introduction Verizon has used advanced analytics methods to reduce the cost of deploying and of operating the FTTP (Fiber To The Premises) network. Two of these are discussed here. The first involves the design of the FTTP network to optimize the network cost. The second is an investigation of reducing the capital cost of the network through reuse of idle facilities. 2. FTTP Overlay Network Design Verizon has deployed an FTTP network, and a large part of it is an overlay of the copper network. Initially, the voice service on the FTTP network was from the existing TDM (Time Division Multiplex) switch. Because of wire center and rate center restrictions in provisioning systems for voice, the initial deployments could not cross wire center boundaries. The fiber followed the copper network from each CO (central office) and each wire center was deployed separately. There is no reason that the economics involved in the design of the copper network would result in an optimal design for the FTTP network by following the copper network. The major consideration is the trade-off between the cost of the CO equipment and the cost of the fiber network. There are economies of scale in the CO equipment to be had by covering a larger area requiring more PONs (Passive Optical Network). However, the larger the serving area is from a host central office, the longer the average PON length, and so the more costly the feeder. Also, for longer reaches, the splitter ratio may need to be smaller requiring more PONs. This reduction in splitter ratio increases the feeder cost and it also reduces the utilization of the CO equipment in terms of premises per PON. To take advantage of the economies of scale in CO equipment, Verizon investigated “CO Bypass” in which several wire centers would be served out of a single central office. In many cases, it was clear that this was more economical, so some of the system restrictions were removed so wire center boundaries and rate center boundaries could be crossed. To explore the potential further, Verizon developed an optimization methodology that, given a cluster of wire centers, selects the number and locations of COs for optimal host locations. The initial input data for the methodology are census block group boundaries divided along wire center boundaries to develop the population distribution within each wire center. Each of the block groups or sub-block groups were treated as distribution areas having the household count from the census data. See figure 1. To determine the cost to serve these areas, cost models of the feeder plant were developed. The number of PONs required for each distribution area from each possible host CO were calculated and the feeder cost from each possible host CO to each distribution area were then calculated. It was not necessary to calculate the cost of the distribution network as this was assumed to be the same regardless of which host CO was serving it. The second cost model required was the CO equipment cost as a function of the number of PONs. There is a fixed cost for introducing FTTP equipment into a CO, and also some economies of scale in equipment utilization. A detailed model was built that took into account each type of equipment required. It included the cost of the basic unit and the cost of expansion modules or cards. This model 978-1-55752-962-6/13/$31.00 ©2013 Optical Society of America NTu2J.4.pdf OFC/NFOEC Technical Digest © 2013 OSA was too complex to incorporate into the actual optimization routine, so a linear approximation was used for exploratory optimizations. When some scenarios were settled on, the more complex model was used to refine the results. The optimization was formulated as a linear integer location/allocation problem  and was implemented within MapInfo, a GIS (Geographical Information System) platform. The optimization engine used was the Frontline MIP (Mixed Integer Programming) Solver . The optimization was used on several clusters to demonstrate the potential benefits of CO bypass. See figure 2. It also was used on partially deployed wire center clusters to determine the minimal cost plan for further deployment. In some cases, local constraints such as lack of CO space, insufficient power, no conduit space left, etc. prevented the minimal cost plan. In these cases, constraints were added to the optimization to find the next best solution. The typical curve of feeder and CO equipment cost vs. number of hosts is somewhat flat in the vicinity of the optimal, so in general these constraints did not increase the costs a great amount. Typically, the cost decreases with more bypassed COs and fewer host COs in a cluster until distances require 1x16 splitters rather than 1x32. The cost then increases as more 1x16 splitters are required. On average, over a large number of clusters, the savings was on the order of 11%. CO bypass became and is the design of choice – justification is required for not doing bypass, where in the beginning, justification was required for doing CO bypass. The system restrictions have been eliminated. Improved optics and improved network energy budgets have increased the reach for each splitter level and these make bypass of larger wire centers economical. Housing Unit Density Housing Units per Square Mile 4,000 to 20,300 2,000 to 4,000 800 to 2,000 100 to 800 0 to 100 (49) (49) (49) (35) (60) Figure 1. Cluster with 6 COs showing household density Figure 2.. Optimal 3 CO host solution for cluster 3. Network Idle Facilities Reuse After the FTTP network had been deployed for a while, the number of idle ONTs (Optical Network Terminals) began to increase. This is a natural result of subscriber churn. When a subscriber discontinues service, the ONT, the distribution fiber, and the splitter port become idle, and the utilization of the feeder, the OLT (Optical Line Terminal) card and upstream equipment is reduced. Over an extended period of time, as most households have or have had FTTP at some time, most households will have an ONT, a drop, distribution fiber, and a splitter port, and also will be using up feeder and OLT capacity. In the long run there may be enough capital equipment placed to serve 90+% of the households in the distribution area, but only 40-50% may be active at any given time. See figure 3. Verizon has investigated methods for reusing the idle capital equipment before too much has been placed, with a goal of having just enough to support the active subscribers. Retrieving idle ONTs for reuse was determined to be too expensive so the focus is on the splitter port and all equipment upstream. The approach is to disconnect the idle ONT at the splitter port, and reuse the port for a new subscriber, and so reuse the capacity of the feeder and OLT. This reuse delays or eliminates the need to place additional splitters and their associated feeder and OLT card. The approach has the potential to reduce the capital equipment required in the long run from the amount required for 90+% of the households to that of around 60%. See figure 4. However, this reuse comes with a cost. First, there is the extra labor of disconnecting the inactive distribution fibers to make ports available for reuse. Whenever a port is made available for reuse, it creates an “orphan” ONT. Requests for FTTP service from households with idle or orphan ONTs are very NTu2J.4.pdf OFC/NFOEC Technical Digest © 2013 OSA common. Reinstalling FTTP to these households requires the additional labor of a trip to the hub to place the distribution fiber into a port so service can be reinstated. Under reuse, reconnecting this orphan will require a disconnect creating another orphan. There will eventually be a large pool of orphans that require reconnections. The reconnection rates vary with the length of time the ONT has been idle and the rate slows as the length of time become longer. Because of this, a reasonable reuse policy is to not reuse ports until they have been idle for some time to minimize the reconnect expense. In order to estimate the potential capital savings and the resulting increase in labor expenses, a Monte Carlo simulation model  of the activities in a distribution area was built using the Risk Solver software from Frontline . This model simulates the life of a hub over 15 years. It has parameters for connect rates for households with no ONT, churn rates for active customers, and reconnect rates for idle ONTs by age. It simulates for differing size hubs (households in the serving area), and uses the distribution of Verizon serving area sizes to generate average results per hub. It keeps track of the inventories by month of households in each status category of no ONT, active ONT, idle ONT, and orphaned ONT. The active plus idle count triggers the placement of additional splitters when needed, and this provides the information necessary to calculate capital required. It also keeps counts by month of the changes in status of: new connections, deactivations, disconnects for reuse, idles reconnected and orphans reconnected. Each of these changes of status requires labor expense, and so provides the information required to calculate expenses. Using this model, the capital saved and additional expenses incurred were calculated for a variety of reuse policies. Hub Evolution - Reuse Starts at 40% HH with ONTs 100% 100% 90% 90% 80% 80% 70% 60% Connected Active LIF 50% 40% 30% 20% 10% Percent of Households Percent of Households Hub Evolution -- No Reuse 70% Connected Active LIF Orphans 60% 50% 40% 30% 20% 10% 0% 0% 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Years in Service Figure 3. Percent connected with no reuse. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Years in Service Figure 4. Percent connected with reuse. A minimal expense policy for making ports available for reuse is to have the installation technician disconnect an idle port as needed for an installation. To implement this policy requires careful controls and good systems to prevent disconnecting a live customer. Other policies investigated would do “batch” disconnects. This allows more control, but creates additional orphans before the ports are needed. Because some of these ports would need to be reconnected before the port was reused, the expenses would be increased. Verizon is currently working on systems issues and methods for operations to implement facility reuse. The potential capital benefits are large, but the systems and operations are not trivial. In both of the cases discussed here, advanced analytical methods provided insight and pointed the way to better decisions which would not have been obvious without these tools. In the first case, integer programming methods were used to guide the network design, and in the second, a complex Monte Carlo simulation model brought to light the level of additional expense required to reuse capital. 4. References  F. S. Hillier and G. J. Lieberman, Introduction to Operations Research, 8th Edition (McGraw-Hill, New York, NY 2005).  Frontline Systems, Inc, Incline Village, NV, www.solver.com.