Contemplating the 5G-enabled Future – Where Will MEC Take Us?

Before Looking Forward, Let’s Take a Quick Look Back

Given the fast-moving nature of our industry, it is interesting, to me at least, to pause for a moment, think back to where we were five, 10, or 20 years ago, and reflect on how far we have collectively come, both from a technological and a societal perspective. 20 years ago, I started a new job that was mainly home based, an experience many office workers are facing for the first time now in 2020.

In the year 2000, DSL was initially being rolled out, but it hadn’t hit my town yet – so very different from the highly connected world we live in today. Roll forward five years to 2005 and technologies we take for granted today were becoming widespread, such as in-home WiFi and DSL, but we still used mobile phones primarily for voice calls.

The first iPhone heralded the rapid rise of smartphones, but it was only released on June 29, 2007, so only 13 years ago – the iPhone has only just become a teenager! That first iPhone came in two models, one with 4 GB of memory and one with 8 GB, revolutionary at the time but an unimaginably small amount of memory now. Technology and our ways of living and working have evolved rapidly since then, and we are now about to start our journey to the highly connected world of 5G. In 10 years’ time we’ll look back on the technology of today in the same way we look back to those pre-smartphone days, when a phone was mainly just a phone, and we’ll wonder how on earth we ever survived without the killer apps of 2030.

Data Center Evolution

This massive evolution in the way we live and work has been underpinned by many technological advances from components/chips, user devices, and telecom/datacom networks, including our use of data centers. In recent years we’ve seen a move toward mega data centers with colossal compute and storage resources centralized in purpose-built facilities that can now exceed over a million square feet. The largest currently, when measured by physical size of the data center, is China Mobile’s Inner Mongolia Information Park, which is 10,763,910 square feet, or just under 1 square kilometer. Mega data centers bring many benefits, with their centralized compute and storage facilities, and many of the applications we use every day wouldn’t be economically viable without the trend to ever larger mega data centers.

5G Changing the Game

5G, though, brings a new twist to how we consider data center infrastructure and where processing and data storage for certain applications will take place. This is particularly important for the new ultra-low-latency services that 5G will bring, which have the potential to fundamentally change the way we live, work, and travel over the next decade or so.

So, let’s initially take a look at mobile network latency in current 4G Long Term Evolution (LTE) mobile networks. In theory, 4G networks deliver services with a round-trip latency of around 10 milliseconds (ms). 5G has a goal of reducing this to 1 ms for latency-sensitive applications. In reality, latency varies considerably in most 4G LTE networks today, with average latency of around 40 to 50 ms and a range of around 20 ms to as high as 250 ms depending on a range of factors such as packet size, distance from the radio antenna, distance between the cell tower and core, network load conditions, etc. While the stated goal of 5G is to reduce this latency to under 1 ms, most network operators seem to be focusing on around 4 to 5 ms for initial 5G low-latency service use cases. To achieve this, the industry needs to look at all aspects of latency, from the end device to the data processing in the core.

Optical Fiber Latency Basics

The optical transport network has a big role to play in this end-to-end latency. Building low-latency transport hardware has always been a focus for many of the vendors producing products for these networks, and this will continue as portfolios evolve for 5G. But to enable the overall network to achieve the lower latency requirements of 5G, we also need to look at the impact of the optical fiber in the network. Light in a vacuum travels at 299,792,458 meters per second, which equates to a latency of 3.33 microseconds (µs) per kilometer (km). Light travels slower in fiber due to the fiber’s refractive index, which increases the latency to approximately 5 µs per km.

So, while we are using the current generation of optical fibers, there is a limit to how low we can drive latency. A 50 km link has a fiber latency of 250 µs, a 200 km link has a fiber latency of 1 ms, and a 1,000 km link has a fiber latency of 5 ms. You can clearly see that to achieve the low latency goal for 5G, even if we assume there is zero latency anywhere else in the network, then we have a limit of around 800 to 1,000 km for initial 4 to 5 ms low-latency services and 200 km for the ultimate goal of sub-1 ms latency. As these figures are for the round trip from the device to the core and back again, the actual distance between these locations will be half of these figures. Of course, in practice, a lot of this budget will get eaten up by other parts of the network and data processing, so these distances will be significantly lower still.

Introducing MEC

To achieve this lower latency, we must migrate to a multi-access edge compute (MEC) environment where the processing of data for latency-sensitive applications is moved to smaller edge compute locations nearer to the end user. Compute and storage resources will be more expensive here, so these facilities will only be used for services that require this level of performance, and there will be an economic balancing act between the additional revenues these services can achieve vs. additional deployment and operational costs.

Operators are typically planning their MEC-based infrastructure to deliver services on the order of sub-4 to 5 ms performance as an initial step, and once these prove themselves technically and economically viable, then operators will evaluate extending this closer to the customer to support even lower-latency services. It is very likely that over the next decade we’ll see a gradual move of dedicated low-latency mini data center facilities closer and closer to end-user devices as this happens.

Impact on Mobile Transport Networks

Mobile transport networks already need to manage the division of the previous 4G backhaul network into front-, mid-, and backhaul domains to support the radio unit (RU), distributed unit (DU), and centralized unit (CU) architecture. We will also have to add MEC locations into this transport domain, which may or may not be colocated with the DU or CU. In a similar way to network slicing, this largely impacts SDN-based control and orchestration software, which needs to manage this environment with optical transport and IP resources via multi-layer, multi-domain, and multi-vendor capabilities along with specific applications to support MEC environments. These advanced network orchestration platforms are starting to be deployed in mobile networks today, and work is underway to test the specific capabilities needed to support MEC in the transport domain.

Looking Forward to Looking Back

So, in a decade’s time, when 5G is well established and is supporting a wide variety of aspects of our daily lives that probably seem unimaginable now, we’ll look back at the technology of today in the same way as we now look back at that pre-smartphone era. The killer apps that will be widespread in 10 years are still to be determined, although autonomous vehicles with V2X communications underpinned by 5G, remote surgery supported by low-latency “tactile” communication, and augmented/virtual reality are often touted as candidates. All these will require many technological advances to become viable and widely used in society, and MEC will be quietly in the background, playing a critical role in enabling these lower-latency applications.