5G is more than just higher download speeds for phones. It's also about latency, reliability, and machine-to-machine communications that edge networks must handle.
By now, we’ve all become familiar with 5G's promise of data rates, with peak speeds of over 10 Gbits/s. This is commonly referred to as eMBB (enhanced mobile broadband) and includes today's familiar internet services such as email, web browsing, and video streaming. But, 5G is more than that and the wired networks will have to adapt to handle the increased traffic and low latency we hear about.
5G includes massive machine type communication (mMTC). This service is focused on limited bandwidth access of remote sensing devices, such as gas, water, and electric meters. Another service is ultra reliable low latency communication (URLLC). URLLC promises to deliver new services where low latency is critical, including remote surgery, autonomous vehicles, factory automation, and numerous tactile applications. URLLC is poised to disrupt the network edge, changing the fundamental network topologies we are familiar with for internet service.
To understand how network topology will change, you need to understand how the network topology is structured for the common internet services we receive today. For illustration purposes, Figure 1 is a high-level network diagram for Comcast, one of the largest Internet Service Providers in the United States.
Figure 1. A high-level diagram of the Comcast network coverage areas for internet access (Source: Comcast)
When we watch a movie over the internet or access email, the traffic we receive is delivered to an ISP at a specific switching node called an internet exchange point (IXP). Those points are identified as the white diamonds on the Comcast map above alongside a city name. The data is routed through levels of the network hierarchy until it reaches a home or business. For uploaded data, the reverse occurs until the data is sent out toward a destination from the IXP.
This traversing of data across the ISP network has been the focus of the net neutrality debate. The net neutrality rules, now revoked, demanded that ISPs treat traffic traversing their network the same regardless of the content, source, or application. The rules did not apply to the internet networks connecting to the IXPs, such as proprietary content delivery networks (CDNs) from providers like Netflix or Apple TV, or Tier 1 internet suppliers such as Cogent or Level 3 that deliver internet data over long distances, connecting between ISPs.
This column has no intent to wade into the net neutrality debate; it was brought up to illustrate how entrenched the standard model of an ISP is, all the way to the last mile delivery of data. This model now faces disruption due to low latency services from 5G URLLC and similar wireline services.
Low latency vs. high bandwidth
We commonly think of network speed in terms of Mbits/s or Gbits/s, a bandwidth metric and a valuable parameter when evaluating networks for delivering high-bandwidth services such as video streaming. Video now represents a majority of all downstream internet traffic globally. At peak times Netflix and YouTube together may represent more than half of all traffic in the United States. In some ways, every evening the US internet becomes a nationwide cable TV system. This fact has led ISPs to create networks optimized for video streaming applications.
What's not critical in these applications is the latency in delivering the content. These applications are fairly insensitive to the total delay across the network, as long as they can buffer the data near the end points and the total bandwidth of the channel is adequate. In general, few people care if their downloaded movie over Apple TV starts two seconds later. This is also true of other services, such as email, file transfers, or data backups.
There is, however, a plethora of promising new services that are latency sensitive. Studies have shown that latency needs to be reduced to approximately a single millisecond for a user of a haptic application, such as playing an instrument or performing remote surgery, to perceive the application as simultaneous.
This is also true of some factory automation applications where there is a control loop requiring tight timing. The approximate latency times of several applications are shown in Figure 2.
Figure 2. End-to-end latency requirements and reliability requirements for several latency-sensitive applications (Source: Ericsson)
5G URLLC devises a coding and scheduling protocol that guarantees a maximum latency of under one millisecond between a user and a 5G base station. It also guarantees very reliable transmission, but we will ignore that and focus on latency.
Regardless of the performance of the 5G URLLC air link, the network topology itself is an impediment. Looking at the Comcast map, there are large distances between IXP points. I live in a mountain town in northwest Colorado that is approximately half way between Denver and Salt Lake City. The distance itself, constrained by the speed of light, is an impediment. The speed of light in an optical fiber is approximately two thirds of its free space speed, or about 200,000 kilometers per second. Consider an end-to-end latency requirement of one millisecond. This equates to 200 kilometers, or about 120 miles round trip. This lowers the service area to 60 miles one-way from the IXP. This does not account for the network equipment latency due to multiple switches and routers directing the traffic through the network hierarchy, which is also significant. Somewhere there has to be the application server as well. Altogether, this compels that the entire application be located as close to the edge of the network as possible. The standard model of data traversing across an ISP network from an IXP to the user is disrupted. A new model of embedding customer equipment into the network edge is required.
Some other factors
5G URLLC and similar low latency delivery vehicles are not the only forces disrupting the edge. One of the very first 5G applications is fixed wireless access, where internet service is delivered wirelessly to the home. Verizon has announced such networks for Sacramento, Houston, Indianapolis, and Los Angeles. This service is a direct competitor to internet services delivered by DSL or cable, challenging the defacto duopoly in many parts of the country. The added competition will change how network providers, consumers, and regulators will look at the network.
The standard model was never pure to begin with. Comcast, by building out its own internal network into a regional network, should be considered a Tier 1 supplier in the United States as well as an ISP. This offers Comcast settlement-free peering with other Tier 1 suppliers, lowering its internet access costs.
Starting in the 1990s, what we currently consider the internet revolutionized long distance communication. The network edge, though getting continuously faster, kept the same topology and function. Due to the forces of 5G and low latency applications, the edge will be soon be disrupted as well.
—Larry Desjardin is a regular contributor to EDN's Test Cafe. He served in several R&D and executive management positions with Hewlett-Packard and Agilent Technologies, and currently manages a consulting company, Modular Methods.
- 5G: Out of the Lab and Onto the Street
- Could local fees kill 5G?
- Five unintended benefits of 5G
- The first killer app in 5G wireless may not even be mobile
- Brooklyn 5G: mmWave more promising than first thought
- Video killed the Energy Star: Why 5G must use less power
- Want 5G? It’s going to take an IP anyhaul overhaul