Residences' move from viewing broadcast TV to Internet TV and businesses' pervasive use of rich multimedia content and cloud services are forcing service providers to re-architect their metro networks. The resulting network designs are bringing network content and intelligence closer to end-users.
Fixed broadband and subscription TV services are the primary sources of residential bandwidth requirements. The home's bandwidth requirement is growing steadily as each resident uses more devices and as more big-screen HDTVs are purchased. Within the access network this steady bandwidth requirement increase is easily met using VDSL2 and fiber-optic technologies. However, the trend to move from traditional TV viewing which employs multicast addressing to Internet TV which employs unicast addressing is fundamentally changing bandwidth requirements on the aggregation network.
This usage shift can best be illustrated by a simple thought experiment. A typical video serving office in a large metro area may serve 100,000 households. According to the Nielsen Company the average U.S. household currently watches 7.5 hours of Internet TV and 155.5 hours of traditional TV per month. Today, that programming is sourced from content stored at the metro-core node. A 35 Gbps data stream between the video serving office and the metro-core node is needed to supply the Internet TV content and about one Gbps is needed to supply the traditional TV content. Now assume that in five years the average household watches 7.5 hours of traditional TV and 155.5 hours of Internet TV per month. Then bandwidth of more than one Tbps is needed to supply the Internet TV content while one Gbps remains as the bandwidth requirement for traditional TV content. Though it may take more than five years to achieve a complete reversal of viewing behavior all market analysis point in this direction. I recently completed an analysis that predicts 31 percent annual growth over the next five years in residential bandwidth requirements on the metro network.
Rich media content also is driving up business bandwidth requirements at double digit growth rates. Like residences, businesses are making heavy use of rich media web browsing and streaming. I developed a bandwidth calculator that predicts the WAN bandwidth requirements for a business establishment (site). It shows, for example, that a headquarters site with 1,000 users will require WAN capacity of 160 Mbps for web browsing (including embedded videos), 170 Mbps for file transfer and backup, 220 Mbps for desktop videoconferencing, and 55 Mbps for social networking. All other applications including podcasts, video conferencing rooms, voice, bring your own device (BYOD), e-mail, and cloud services require an additional 119 Mbps for a total WAN requirement of 724 Mbps. Under the networking and computing models of ten years ago this large site would have been served by a 45 Mbps WAN connection.
It is interesting that some of the most mission critical business applications such as voice, e-mail, and cloud services do not require that much bandwidth. These types of applications, however, impose additional security, performance (latency), and availability constraints on the metro network design.
The metro network must be redesigned to cost effectively meet these new requirements. The traditional design was optimized to meet the "long-tail" hypothesis that each of us would use unique content located anywhere in the world. As such end-offices and aggregation nodes funneled traffic to a metro-core node where network services were provided and a connection was made to the long-haul network. This traditional design assumed that 20 percent of the traffic stayed within the metro area while 80 percent traversed the long-haul network. Over the last few years the large video traffic flows have driven network operators to cache and/or source video content at the metro-core node so as to reduce the cost of the long-haul network.
The new metro design recognizes the need to invert traffic flows so content and move network intelligence closer to the end-users—"eyeballs". My estimate is that 50 percent of the traffic will stay within the area served by each aggregation node, 30 percent will flow among data centers (or carrier hotels) located at the aggregation and metro-core nodes, and 20 percent will flow over the long-haul network. This will involve collaboration among many service providers including network operators, cloud services providers, content distribution networks, mobile operators, and wholesale carriers. This new design reduces cost by substituting virtualized processing and storage resources for network transport resources.
The new design also envisions increased network intelligence using tools such as software defined network (SDN) and network function virtualization (NFV). This intelligence can be distributed across the metro area at low cost through centralized control and virtualization. It enables substantial cost reductions (especially in operations expense), increased service velocity, and improved service creation. Specific examples include virtualized transparent caching, virtualized video transcoding, multiscreen cloud DVR, virtualized CDN, multilayer SDN, and virtualized business and residential CPE.
The massive growth in both residential and business bandwidth requirements cannot be met alone by moving content and intelligence closer to the end users. This action primarily reduces network transport capacity requirements and the associated capital expense. Seven to eight operations expense dollars are incurred for every capital expense dollar. SDN and NFV initiatives, however, have the potential to significantly reduce operations expense by making networks programmable. Programmable networks require many fewer manual work activities than traditional networks. This reduces the high cost labor component of delivering network services. Also, it frees up scarce networking talent to pursue value-added revenue initiatives.