The Coronavirus pandemic represents the biggest test for the Internet in living memory. Never have there been so many people dependent on it for so much. In the space of mere weeks, it has transformed physical classrooms into virtual ones and enabled real-time, business-wide collaboration from the home.
Indeed, the profound changes to the way in which we live, work and play are destined to open a Pandora’s box of questions: do we need to travel vast distances to sit at a desk or to meet face-to-face when many tasks can be fulfilled at the touch of a fingertip?
It will, of course, be the Internet’s ability to adapt to these sudden changes in demand that will influence the answer to the above questions. In this article, we will examine how the profile of Internet usage has changed radically in recent times, and how those changes are being supported by the telecoms networks that prop up the Internet.
The profile of Internet usage: Changes in geography, time of day and peak load
In seeking to understand how the profile of Internet usage has undergone a tectonic shift, there are three key dimensions in which to focus. The extent of change witnessed within each of these three dimensions varies on a country-by-country basis, but the factors that have triggered the changes are universal.
The most obvious dimension in which Internet usage has changed is the geographical distribution and mobility of users. Government-led lockdowns have become progressively more stringent across the world, confining billions of people to their homes for a large proportion of the day.
A mass exodus of the population from ultra-urban areas, business districts and education campuses has never been seen before on this scale. It has, in effect, pushed the population out into and beyond suburbs, exurbs and rural areas, where the vast majority of people have their homes.
This means that an overwhelming share of Internet traffic is now generated in residential areas, and on the fixed and wireless networks that serve those areas. Meanwhile, usage has plummeted on enterprise and education networks, and the same is true at transport hubs and along transport routes.
By virtue of the above trends, the competency of the “last mile” has become a key consideration. Whether it is a copper, fibre or wireless service that forms the access network, congestion has increased as people cluster within households at the network edge. The shared resource here will be, for example, the spectrum deployed (in the case of FWA or in-home Wi-Fi).
The second dimension in which the profile of Internet usage has changed is the time of day. There has been an extension or elongation of peak traffic load during the traditional busy hour in the evening. Put simply, demand is being spread out across the day more evenly than before, reducing the magnitude of the delta between peak and non-peak hours.
Remote working and remote education, along with increased consumption of streaming services during daytime hours, are key drivers of this change. As a result, traffic has increased considerably in the late morning and during the course of the afternoon. Early morning traffic, however, has declined as people start their workday later at home.
Collaboration tools such as Microsoft Teams (which saw video call usage jump by more than 1,000% in March) and Zoom (which saw daily meeting participants balloon from ten to two hundred million between December and March) are now a fixture of the workday.
In addition to the spreading of Internet usage across earlier hours in the day, it has also increased after midnight. This is notable because traffic tends to fall off very rapidly when people reach their bedtimes. The pandemic is seeing people staying up later and accessing services such as Netflix and online gaming as they are not working early in the morning.
Finally, the third dimension in which the profile of Internet usage has changed is the volume of usage itself. Not only has the busy hour extended, but its peak has been pushed up, breaking records in a number of countries. This is why core network and application server capacity has been strained in recent times.
Data from Internet exchange points (IXPs), where many operators and content providers meet to exchange traffic, can be considered to be indicative of overall Internet usage. It is important to note, however, that this data does not provide a perfect picture of residential traffic because it is blended with business traffic. Furthermore, many large content providers peer within the core network of operators, effectively avoiding IXPs.
Nevertheless, IXPs have been reporting some fascinating trends. INEX in Ireland and DE-CIX Frankfurt in Germany, for example, witnessed average traffic volumes grow by 30% and 10% respectively during March. The London Internet Exchange (LINX) has reported similar increases, and INEX broke its all-time traffic peak several times in March (reaching more than 500Gbps).
The response to the changing profile of Internet usage is multi-layered
In response to the extraordinary circumstances presented by the pandemic and its impact on the profile of Internet usage, significant efforts have been made at all levels to maintain quality of service (QoS). Coordination and cooperation between the multiple layers that make up the Internet is key to address issues proactively.
The aforementioned IXPs have responded to the increased demand by enhancing network capacity. INEX has boosted core capacity by 30% and provisioned 30% extra port capacity for its members. This kind of move is important because, as the interconnection ports of the Internet, IXPs enable traffic to move around the world cost-effectively and efficiently.
Some of the largest content providers for video and gaming are playing their part too. On-demand video, after all, represents a disproportionately large share of Internet traffic, and the pandemic is sure to amplify that trend. Services such as Netflix, YouTube, Disney+, Apple TV+ and Google Stadia have all announced temporary adjustments to reduce their bandwidth usage.
Of course, fixed and mobile operators cannot be forgotten either. Notably, Opensignal has reported a marked increase in the time spent on Wi-Fi across most markets where restrictions on the movement of people have been implemented. In Spain, for example, a week-on-week increase of more than 25% has been noted.
However, at the same time, mobile networks have been experiencing unprecedented demand in many markets. Mobile voice traffic, in particular, has exploded despite a backdrop of year-on-year declines. In the US, Verizon reported a 15% increase in call durations, while AT&T saw voice calls shoot up by 44%, over different periods in March.
The increased usage and the concentration of the population in areas where site grid density tends to be lower has resulted in a fall in median average weekly 4G LTE download speeds, according to Opensignal. Markets such as the UK and Italy have seen a speed decrease of more than 15% and 20% respectively.
In order to maintain performance, a number of strategies have been employed by mobile operators. In Ireland, as detailed in our previous article, the telecoms regulator (ComReg) has released additional spectrum in the 700, 2100 and 2600 MHz bands on a temporary basis, subject only to a nominal per-license fee of €100.
Separately, Dish Network in the US is lending some of its unused spectrum to mobile operators there at no cost. This gesture of goodwill began when it gave T-Mobile access to the full breadth of its 600 MHz portfolio, and has since been extended to AT&T with 20 MHz of AWS-4 (Band 66) and the entirety of its 700 MHz holdings.
Some regulators and mobile operators have also been encouraging consumers to call with Wi-Fi, traditional landlines or OTT services such as Skype and Whatsapp. This is so as to reduce the strain on the cellular voice network while maintaining a high QoS.
The capability of mobile networks that are dependent on a circuit-switched fallback (CSFB) architecture to cater to a large volume of voice calls has been somewhat reduced by the refarming of spectrum in recent years.
Conclusion: A paradigm shift in how we think about and use the Internet
The second-order effects of the coronavirus pandemic are heralding a paradigm shift in how we think about and use the Internet. It is has become a centrality in all aspects of society, touching everything from work to education and entertainment.
A multi-layered and coordinated response between the layers that make up the Internet has ensured that we can continue to stay connected, even as the profile of usage undergoes significant changes across the dimensions of geography, time of day and peak load.