In networking, as in all businesses, everything comes down to the spread between cost and price. Since the early 1980s, optical advances have driven down the cost of capacity by driving up the number of bits that can be sent along a path. It's hard to believe it today, but only 15 years ago, the average corporate headquarters site had less access bandwidth available than a consumer with fibre-based Internet access could buy today for about $150 per month. Optics has changed everything.
Despite the revolution that network optics has created, network planners still sometimes fail to fully factor in the impacts of optical trends. While this rarely leads to outright project failures, it often leads to deploying less-than-optimal solutions for increasingly complex problems and opportunities.
It would be fair to say that the most significant technology planning challenge today is the effective use of optical networking. The three areas of optical impact most likely to be problematic are:
- The PON revolution in access
- The metro bit gradient
- The new role of the core network.
Up to now, optical improvements have been felt deep in the network where traffic could be concentrated to multi-gigabit speeds. Passive optical networking (PON) has brought optical power directly to access and outside plant, increasing the potential connection speed for consumers by 1,000% or more. PON has evolved from an early version based on ATM (called APON then, now renamed to broadband or BPON), to today's dominant gigabit PON (GPON) and a future 10G-PON, and an Ethernet or EPON. All of these technologies create an optically spliced multi-site connection path that reduces per-user cost by largely eliminating electrical devices in the access connection.
PON planning challenges
The challenge that PON poses to planners is the "current -versus -opportunity" -cost. A PON fibre connection can support gigabits of transmission to each home or business site without being replaced, where an access connection based on electrical technology like ADSL or VDSL requires regular maintenance and replacement and must be upgraded in the outside plant (the remote) to take advantage of higher-performance standards.
There are also basic physical limits on these electrical/copper hybrid access architectures in terms of maximum capacity. But PON today will cost between four and 10 times the cost of DSL to "pass" a customer. Will the bandwidth headroom it offers in the future justify its cost in the present? The answer probably lies in the economic density of the area to be served. The more dollars of potential revenue are concentrated in a given area, the more likely PON is the best way to serve it.
Changes in metro area traffic
The second problem area for planners is the sharp change in traffic density in the metro network relative to the core. This "bit gradient" is exacerbated by the delivery of commercial content like IPTV, which is normally served from a metro service centre in each major population zone and not centrally distributed through core network connections. Three hundred HDTV channels could require as much as 2.4 Gbps of programming delivered to every central office, and yet would not necessarily generate any "core" traffic at all. Today's IPTV schemes use IP multicasting and thus require IP devices in the metro network.
The problem is that personalised or video-on-demand (VoD) services could totally change the requirements for multicasting TV channels and substitute personal video streams for broadcasting. Modelling today's TV habits, a central office (CO) with a 20,000 household service area would likely require about 64 channels to be delivered, which is 0.5 Gbps if they were broadcast. If the same population of users simply consumed personalised streams of HD, the users would require 160 Gbps, a 300-times increase.
PON deployment could have a dramatic impact as well. PON in the access network permits broadcast channels to be sent using traditional RF-over-fibre techniques, so no actual data traffic is generated at all. Neglecting Internet traffic, such a video strategy would generate no metro load at all. But if VoD in a data-delivered form became the exclusive video service, it would again generate a 160 Gbps load per CO.
These variables show that the metro infrastructure will have to contend with enormous potential shifts in traffic, which argues strongly for an infrastructure that is rich in agile optics, including ROADMs, and also one that uses DWDM heavily to create a large capacity pool for allocation.
Metro traffic effects on the core
The traffic variables in the metro picture raise the third point, which is the role of the core network in the whole process. A transition of users to video-on-demand entertainment would also make it more likely that specialised content would be required for a given CO, content that would not likely be stored within the metro area. Again forgetting Internet traffic for the moment, the core traffic for the local broadcast entertainment approaches above would be zero. And yet in theory, a large percentage of the 160 Gbps of per-CO traffic from personalised entertainment could be fulfilled from outside the metro area, and thus it would generate core traffic.
The question of where that incremental traffic would then originate becomes the key question for core planning. If we assume that the content everyone wants delivered is the content that someone will pay for, then the critical question for core design may well be the question of the advertising paradigms that would be successful in sponsoring content delivery. If there are paradigms that are so flexible they can even accommodate user-generated content, then core traffic could end up distributed as broadly as it is today. However, while there is considerable viewer interest in free user-generated content, however, it is not clear whether there is any way that delivering this content could be linked to a business case. Thus it is not clear that building out infrastructure to support it would be a prudent decision.
Based on current user behaviour, it is likely that a few central points would serve "long-tailed" commercial content to metro areas and that content delivery networks (CDNs) would shortstop traffic from the core. In these cases, it might well be most efficient to simply create an optical mesh of metro networks and call it a "core." In fact, if video traffic were to swamp other forms of traffic in the future, it is very possible that video delivery (via CDNs or other architectures) would create a series of parallel "core-lets," each focused on delivering a specific set of content to a specific set of metro areas.
Meshing the metro areas with optical links is an exercise in DWDM, and it seems likely that the high cost of providing broadband access infrastructure to users will continue to reduce the number of access providers, making optical meshing an increasingly attractive core technology. However, an electrical device (a switch or router) would be needed to couple core routes into the metro network and on to the user. As optical capabilities increase, this hand-off device is likely to become more massive.
Network planning and design is driven by both the changes in optical technology as a driver to equipment change, and by the changes in optical technology as facilitators of new applications and new traffic. That is a significant change from the early days of SONET, and it is also a trend that seems certain not only to continue but to accelerate.