Recently there has been a lot of talks about the next generation cellular wireless communication called 5G. Along with 5G, there are also a lot of new terms being thrown around – NR, Millimeter Waves, Small Cell, Massive MIMO, etc. Many have targeted their focus on the insanely fast speed that 5G can provide (up to 20 Gbps), but I argue in this article, in both technical and practical perspectives, that the latency characteristics of 5G network far outshines the bandwidth increase that it brings. In addition, looking at the current technology that we have, I bring the overhyped excitation down to a realistic level.
First let’s define what latency is. In laymen’s terms, it is the duration between a request being sent and a response being received. If you said “Hi” to a friend, and 5 seconds later they replied “Hello!”, the latency of your communication would be 5 seconds. Why does latency matter? Imagine this situation – you urgently need to get to a meeting, but you don’t have a watch, so you shout across the room to your friend “Hey what time is it?”. However, your friend took a little bit of time to process the question, then a bit more time to get out their watch, and a bit more time to read the time, and 5 minutes later, they said “It’s 11:03AM!”, by which time you are late for your meeting. This is obviously a massive exaggeration, however, since computers process information so quickly, a latency of a second is considered unacceptable in most cases, which is why latency is usually measured in milliseconds, ms). When you type “Good Computers” into Chrome’s search bar, a lot of things are actually happening in the background. There are some optimisations but in simple terms, every single character triggers a web request searching for phrases starting with those:
- “G” – searches for G
- “Go” – searches for Go
- “Goo” – searches for Goo
- “Good” – searches for Good
What this means is that by typing “Good Computers” into Chrome, you had effectively Googled 14 times. However, this isn’t the end of it. Every time you search for something, there are a bunch of requests that need to be done, e.g. DNS lookup, Google Ads services, APIs, chrome history sync, etc. So it ends up being hundreds of requests just for one simple search. Imagine if a single request took a second – would you really want to wait two minutes (assuming they are sent sequentially, which they realistically aren’t) to search for some good computers? Mind you, in day-to-day life, this isn’t that much of a big deal. Google taking a second to load won’t (under most circumstances) affect anyone. However, if you are on systems relying on low latency to function, such as autonomous cars that need to communicate with each other or the cloud for information about the road and traffic, every millisecond counts. If you are driving (or being driven) at highway speed, 110 kph, in 1 ms you will travel 3.1 metres. For city driving at 60 kph, 1 ms = 1.7 metres.
Low latency can enable a huge range of applications, such as cloud gaming – hire a cloud GPU, process the game, and stream it back to your thin device (which might be a TV or a phone) for you to view and control in real time; augmented reality – smart eye wear which identifies information and displays in your vision in real time; thin devices – composed of just a screen and control devices, all else are streamed from the cloud (which Citrix and VMware have been successful in regards to WiFi and LAN devices), enabling our phones to be even thinner and lighter consisting of just the battery and a touch screen with almost endless upgrade capabilities. The possibilities are endless.
5G has a specific target category – Ultra-Reliable Low-Latency Communications (URLLC), targeting to reduce the latency to just 1 ms. This is a monumental achievement if it’s done. We mostly consider WiFi to be low latency devices, and some of the concepts that I mentioned can already be done under WiFi environment (streaming game from your computer to your TV via WiFi, Citrix and VMware solutions). However, the typical latency for WiFi even when you are standing right next to the router is around 5 ms due to the packaging and transmission overhead with the 802.11ac and TCP/IP protocols.
Now speed – 5G NR apparent has a maximum achievable data rate of 20 Gbps. That’s 2.5 GB/s. Even the realistic speed is half that, 1.25 GB/s, what are we doing that requires faster download speed than most SSD’s read speed (read: it’s faster to download something from the internet than it is to fetch locally)? Surely the data cap on our wireless plans would not allow anyone to experience this kind of speed for very long! This is perhaps just me being overcritical living in a country where 4G speed usually exceeds 100 Mbps. When I was travelling in China, the 4G speed would drop to the low teens because of the bandwidth limits of the towers and the limited spectrum efficiency of 4G LTE networks. 5G NR, which utilises massive MIMO, would be able to provide 3 to 4 times the spectrum efficiency of 4G, allowing more devices to be passing through more data concurrently.
What I am arguing in this article, is that although 5G can bring previously-unimaginable possibilities onto the market with its drastic improvement to latency and bandwidth, the surrounding environment (hardware and software developers, data providers) still need to play catch up to the technology. Your business or experience will not be instantly better once you flip the switch to 5G, as some people believe. As far as I remember, previously the communication has always been playing catch up with the content. 2G was very slow but it’s a start, and we wanted faster network experience. 3G was a lot faster, but we still wanted faster network. 4G was once again a lot faster, and we finally feel happy with the speed and the experience that high speed internet brings to our mobile devices. 5G is the first time, in my opinion, that the mobile communication technology has outpaced the demand. It is the first step towards maturity in the market. Back when CPU was massively under-powered and computer resources were limited, programmers had to come up with ingenious ways to cram more into little space. Nowadays, frequent CPU upgrades become unnecessary for most users, as its computational power, under most circumstances, far exceeds what we need. It used to be the case that programmers think in “What can we do given these limitations?”, and now the thinking should be moving towards “What can we do to harness all the power we have at hand?”, or … with great power comes great responsibilities.
5G is coming. What are you going to do with it?