5G Latency — Reality Checks
(Last Updated On: December 9, 2018)
The industry is flooded with a lot of “5G will save the planet” without doing the homework to understand what is going on. It is shocking to see disinformation statements that “5G is expected to slash data transmission delays from about 30 milliseconds to less than one.” Yes! 5G defines the law of physics! 5G will now allow you to download from San Francisco to Singapore across the Pacific ocean in 1 millisecond!
That is an extreme statement to get people to wake up to the realities of the hype. Before I continue, please note that I’m a huge advocate for the evolution of our mobile infrastructure. We’re moving to a world where everything and everyone is “untethered.” But, with +30 years of designing, building, deploying, and operating networks, I understand the hard realities of expectations matching realities. Many of the people who are talking about the benefits of 5G have never done the hard work of deploying 3G or 4G networks. They don’t understand the complexities of RF planning. They have not been on drive test and network optimization exercise working against the clock to get a section of the city ready for service. 5G is going to give the industry options, but it is not a “plug & play” technology. It is not a technology that is going to be deployed everywhere. It is not a technology that is going to save telecommunications companies. The reality is that it will be another evolution, another tool, and an important part of connecting people and things.
Let us start the 5G reality check by looking at “1ms latency.” Do a search of the Internet for new articles with the phrases “5G” and “1 millisecond latency.” You will have a whole range of articles advocating the “latency benefit of 5G.” These authors buy into the vendor hype of “5G improves latency” without digging into where the latency is improved. What both are missing which part of latency is being improved. To help people understand, we’ll walk through a simple illustration of where we see potential 4G/5G latency improvements.
We can break a 4G (and 5G) network into three simple elements of latency:
Air Latency. This is the latency between the user device and the cell tower. For a 4G network, the latency added to data download would be a variable between 10 ms and higher. The variables will depend on the signal strength, the number of devices in the cell sector, the distance, the noise, and many other factors. THESE FACTORS DO NOT GO AWAY WITH 5G!
5G Reality Check — in a 5G network, the aspirations are to have the air latency down to 1ms — 5ms. This is where you get the quotes that “5G is expected to slash data transmission delays from about 30 milliseconds to less than one.” First, note that this is one element to the total latency between the content (a web page) and the mobile device. 5G is not going to allow you to download www.senki.org in 1 millisecond. We still have a lot more latency variables to deal with.
End-to-End Mobile Latency. This is the total latency between the RF unit (eNodeB) and the gateway to the Internet. This includes the wide area network that connects all the cell towers to the core of the network (the Radio Access Network — RAN), all the routers and switches in that network, all the elements in the core data centers (the Enhanced Packet Core — EPC), and then all the elements in the gateway to the Internet. That final “gateway” will include things like the Carrier Grade Network Address Translation (CGNAT), security firewalls, deep packet inspection, and other elements. This latency is also a variable between 30ms and ~100ms (or more). The variable is from the number of network devices, the distances of the RAN’s network, the load of the network (more congestion means deeper queue resulting in more latency/jitter), the load on each element in the network (overloaded elements in the EPC), and the activity of the users. Again, NONE OF THESE VARIABLES DO NOT GO AWAY WITH 5G! If anything, more variables will be added as Carriers add the 5G core, slicing, and other complexity to the network.
5G Reality check. The aspirations for 5G is to have the mobile end-to-end in the 5ms range. That is going to be a challenge with the way networks are built and managed today. There are ways re-architect a 4G/5G network to dramatically cut down the latency. Note the term …. REARCHITECT. This engineering effort cost time and money. That means each Carrier will be mindfully reviewing the business case for 5G before rearchitecting the network. In the meantime, they will look for other options to cut down on the end-to-end latency.
End-to-End Internet Latency. This how fast it takes to download a website to your phone. This is where the major confusing comes in with the benefits of 5G. Internet latency from the Mobile Carrier to the downloaded website is outside of the control of 5G. There are things the Mobile Operator can do to reduce the end-to-end Internet Latency. They can partner with Edge-Compute companies like Akamai Technologies and put these Edge Compute systems integrated with the core of their network. They can peer at Internet Exchange Points (IXPs) and peer with as many Operators/ISPs as possible. They can partner with the Content and Cloud Operators — peering, putting in private network interconnects (PNIs), and other synergies. These are all factors Mobile Operators are investing in with 4G network. Once again THESE VARIABLES WILL STILL APPLY WITH 5G!
The articles in the press are leading people to think there is some sort of magic technology wand that will be waved and “5G will be wonderful.” The 5G realities are different. The benefits of 5G are going to be huge, but gaining those benefits require diligent engineering. Network Architect will rethink how we build the RAN, peering at more IXPs. Radio planning architects to be creative in the new use of “5G New Radio.” Facilities engineers will be doing the hard work to find space for new towers, mini-datacenters (for the cRAN and vRAN equipment), and other equipment. EPC Core architects in the Mobile Operators will be re-tooling for a world of virtualization with SDN, NFV, and other ways to save cost, power, cooling, and (most important) space.
The 5G evolution is not “magic.” We’re not going to get “5G 1 millisecond latency” by waving some special vendor wand. We’ll get there through hard engineering, architecture, planning, and tight collaboration with peers throughout the Internet.
Originally published at www.senki.org on December 9, 2018.