Network strain: The risks and the reality
Tue 7 Apr 2020
Join us on 30 April at 11:00 BST as we host experts from techUK, Vertiv, LINX and 4D Data Centres to discuss how critical infrastructure is coping with the rise in network traffic. Register.
The internet is not about to implode. But ISPs and telcos are prioritising traffic and halting infrastructure upgrades
Separating this pandemic from predecessors is that information technology is firmly on our side. In the absence of sport, Netflix is keeping families entertained. FaceTime or Zoom are connecting us safely from a distance while ordinary socialising presents a risk. And a plethora of cloud-based workplace tools are channelling business productivity from newly-distributed workforces.
Nobody doubted our dependency on connectivity before coronavirus, but the degree to which the thread of the internet holding society together during this crisis is striking. While restaurants, pubs, cinemas, cafes and more turned off their lights, data centres, network operations centres and internet exchanges geared up for an unprecedented rise in traffic.
What has changed?
Attention has been understandably directed towards the internet’s ability to operate as normal in these circumstances. Internet usage stats vary from country to country. But compared to expected trends, over the last few weeks, global internet usage is higher and more sustained, and there has been a shift from business and academic networks to mobile and home broadband networks.
Europe-wide telco Vodafone has seen a 50 percent jump in traffic since the beginning of March. Telecom Italia SpA fixed-line data volume is up 90 percent since Italy went into lockdown. And Openreach, which owns and operates most of the UK’s phone broadband lines, said daytime data consumption has almost doubled over the last three weeks.
Remote workers are probably not adding a great deal to overall broadband traffic. Tasks like sending emails, updating databases and writing reports have minimal data impact. It is rather the combination of a rise in data hungry tasks like video streaming and gaming, occurring in tandem with remote working, that is fuelling the surge.
The dynamic in mobile networks is less consistent. Telecom Italia SpA reported a 30 percent jump in wireless traffic while UK telco BT has seen a 5 percent decrease in mobile data usage. It’s likely that the majority of new traffic is going through home broadband networks instead of wireless networks, but carriers are dealing with a spike in calls as business landlines gather dust. Switzerland telco Swisscom suffered an outage it blamed on an “overload” of calls, while Three in the UK is handling 40 percent more voice calls than last week. “At a time like this, phone calls are only going to become more important as the country and the rest of the world adapt to a new normal,” said Paul Carter, president and CEO at Global Wireless Solutions.
Meanwhile, DE-CIX Frankfurt, one of the many global internet exchanges underpinning the internet, recorded an all-time traffic peak of more than 9.1 Terabits per second.
Cloud providers have not been exactly clear about how cloud demand as a whole has been affected by coronavirus. Microsoft Azure recently retracted a statistic claiming its cloud platform had seen a 775 percent jump in activity in regions enforcing social distancing, while AWS and GCP are quiet (although a source familiar with the matter told Techerati AWS usage is up around 30 percent).
The statistics we do have involve specific cloud-based service usage. Between 12-19 March, use of Microsoft’s Teams group-collaboration software grew from 32 million to 44 million daily active users. Last week, Google Cloud CEO Thomas Kurian revealed Google Meet daily usage is more than 25 times what it was in January.
Need for speed
Fears that the internet may implode are misguided. The internet is a mix of fixed-line and mobile networks run by governments, telcos, ISPs, cloud providers and academic institutions, supported by a constellation of internet exchange points or IXPs. This distributed architecture means there is no danger of a global or even regional internet blackout.
But if bandwidth demand exceeds capacity in parts of the network, it could lead to temporary downtime or reduced download speeds. While that doesn’t sound too dramatic – it would hamper critical workers’ ability to work and halt a country’s newly-remote workforce in its tracks. “We must strive to maintain our critical infrastructure to ensure mass remote working communication equipment remains operational at optimum levels,” said Gabriel Bonilha, EMEA Professional Services Manager at Vertiv, a global critical infrastructure specialist.
IXPs effectively serve as the ports for all of the internet’s networks. They improve the affordability and efficiency of global data transfer, in turn, easing bandwidth pressure on individual networks. If a major IXP suddenly became a bottleneck, all of its interconnected networks would see bandwidth capacity strained, dogging access and download speeds for users.
So far, the rise in data demand has not affected the bandwidth capacity of IXPs. DE-CIX, the world’s second-largest interconnection provider, said even if all companies in Europe operated entirely remotely and UEFA’s marquee football competition was broadcast in parallel, it would still “be able to make the necessary bandwidth available for seamless interconnection”. This is because IXPs like DE-CIX regularly increase bandwidth capacity to manage the yearly growth in data traffic and always leave reserves in the tank if necessary.
ISPs, telcos and cloud providers
Networks routed through IXPs must also be able to handle bandwidth demand to avert localised online gridlock or crippled speeds. Their ability to do so depends on starting bandwidth capacity and their ability to add or allocate more to deal with peak demand rises in certain areas.
Some are more advanced than others. Vodafone, which is reporting that its networks are coping fine with new demands, uses artificial intelligence to automatically assign capacity in busy parts of the network, for instance.
The latest results from speed tester Ookla reveal very few countries have seen decreases in service speed, and ISPs are reassuring customers that capacity is under check for now. Although, operators in Spain and France are signalling that demand is beginning to impact service quality.
When capacity is stretched, traffic can be prioritised so critical bandwidth is protected over less essential services such as video streaming or gaming. ISPs and telcos have the ability to reallocate bandwidth but are understandably reluctant to intervene as service quality is heavily regulated and the norm of net neutrality is widely supported. So far, in Europe, only French operators have explicitly stated that they are working with service providers to allocate bandwidth away from video-streaming platforms as an exceptional measure to reduce peak load traffic.
In a recent blog post, David Belson, senior director of internet research and analysis at the Internet Society, said cloud computing platforms should have the storage, compute and bandwidth capacity to handle increased demand. Recent cloud outage data from ThousandEyes bears this out, and actually indicates a decrease in outages compared to other years.
Nevertheless, Microsoft has said Azure will begin prioritising emergency personnel, organisations working on the frontline of coronavirus efforts, and remote workers using teams. “We are working closely with first responder organizations and critical government agencies to ensure we are prioritizing their unique needs and providing them our fullest support,” the company wrote in a blog post.
The preference is to tackle network strain with a bottom-up approach, but customer assent is by no means guaranteed. Spanish operators released a joint statement calling on customers to save data-intensive tasks for off-peak hours and many others have urged customers to surf sensibly. Bonilha added that if necessary home-workers could ease pressure on home broadband networks by “falling back” on 4G connections.
Individual service providers have an important role to play, too. Video streaming platform Netflix agreed to reduce video bitrates in Europe for 30 days, the effect of which would be a 25 percent decrease in the service’s data consumption during the period, and Disney Plus has enacted similar measures. Meanwhile, Playstation has limited its gaming network’s download speeds. “Prioritising bandwidth will require a group effort, not only from institutions and telcos, but also from people at home,” said Bonilha.
Sitting between commercial networks and services are content delivery networks (CDN), which ease pressure on networks and the services that rely on them by serving cached content from “edge” locations located closer to the user. One of the reasons Netflix is one of the most reliable and highest performing streaming services is that it runs its own purpose-built CDN of “edge nodes”, which handles 90 percent of traffic.
“Think of edge locations like local Amazon warehouses, smaller and local access points that help optimise on delivery. Edge capacity allows network operators to store and retransmit data closer to the final user and in turn this helps ease network delivery to remote locations,” said Bonilha.
Some CDNs are having to temporarily suspend content neutrality, too. One of the largest, Akamai, whose customers include Microsoft and Sony, is reducing gaming software downloads at peak times, completing the downloads at normal speeds late at night. “This will help ensure healthcare workers and first responders working hard to contain the spread of COVID-19 have continual access to the vital digital services they need,” wrote Tom Leighton, the CEO and co-founder of Akamai, in a blog post.
Possibly the biggest impact of the coronavirus so far stems from lockdown measures that have forced operators to limit maintenance and network upgrades. “Currently only critical interventions are being made on the networks as the priority is to keep workers at home,” said Bonilha.
This won’t affect an IXP like DE-CIX, which can monitor and maintain its data centres remotely. ISPs and telcos on the other hand typically have to physically access cabinets dotted across their network to perform maintenance.
Even though telecom and ISP infrastructure employees are widely recognised by governments as critical workers, operators are playing it safe by halting unnecessary maintenance. At the end of March, OpenReach revealed it was taking the unprecedented step of limiting or delaying new broadband and phone installations, repairs and engineer appointments nationwide. The likely long-term effects of all of these restrictions are further delays to many region’s fibre upgrade and 5G rollout plans.