Published: April 2, 2020

Introduction

Novel coronavirus (COVID-19) is perhaps the largest disruptor to the global economy since the 1918 'Spanish flu' pandemic. Fear, uncertainty and police enforcement of precautionary measures have changed the behavior of communities, although with over 100 years of advances in technology, our interconnected society can still conduct business while working from home. The sudden shift surfaces rough edges, although the underlying ability for millions of employees to work remotely for their respective employers' hinges on the cloud – and the ability of cloud platform operators to scale to this unprecedented demand. The profits of cloud providers depend on the accuracy of their predictions, but few would have predicted that

The 451 Take

Enterprises that have built cloud-native, scalable applications in public and hybrid clouds are in a good position to address opportunities or threats as a result of the COVID-19 pandemic. Ultimately, it is public cloud's ability to scale up (and down) that enables enterprise flexibility when the unexpected happens. Present circumstances may act as a catalyst to greater public cloud adoption. However, public cloud providers have ownership of capacity planning, and now is when contingency planning really comes into its own – even the mighty hyperscalers couldn't have forecasted the events of today, and they are likely relying on excess contingency capacity rather than forecasted capacity. Outward signs indicate that they are handling capacity well – there are no obvious signs of performance degradation, downtime or price spikes as a result of scarcity.

Cloud comes into its own for Enterprises

At times like this, the scalability of the cloud really comes into its own. As the behavior of the populace changes as a result of the virus and approaches to its management, we can expect some web applications to explode with traffic, while others might dwindle to near nothing. News websites are likely to spike as a frightened public looks for information; e-commerce sites might receive lower hits (or shutdown altogether) as consumers wait for more optimistic times. Scalable cloud applications should be able to grow and shrink to meet this changing demand automatically. For those suffering from lower hits, scaling down at least saves a bit of cash – a silver lining in challenging times. For those experiencing a surge, the cloud lets websites continue to perform under pressure.

Crucially, this scaling can be configured automatically, with no need to forecast or plan capacity. With personnel challenges as employees contract and recover from the virus, this automation can make a big difference. With public cloud, the cloud provider – not the enterprise – is responsible for capacity. Conversely, many private clouds will sit unutilized in datacenters, while others will be overwhelmed as demand exceeds the steady-state capacity expectation.

Enterprises operating hybrid IT environments can also 'burst' to meet increased capacity demands, but the ability to scale down is limited, given the fixed-capacity nature of their dedicated private cloud resources. Unfortunately, for enterprises that haven't already built scalable cloud-native applications, it is too late. Enterprises that were prepared will be best placed to make the most of tough times. It's the same with business continuity planning – if you haven't planned before the event, your options now are limited. From our recent COVID-19 flash survey, 41% of respondents are already experiencing increased strain on IT resources. However, 91% state that their IT systems are equipped to deal with the fallout of the pandemic.

COVID-19 might be the catalyst for enterprises to look beyond their own four walls, and into public, hybrid and multicloud options as they face the reality that an event like this may happen again.

Cloud Providers Face Challenges and Opportunities 

Within public cloud, the cloud provider – not the enterprise – is responsible for capacity. Thus, the cloud provider's biggest planning exercise is the provision of capacity – not just the electricity, connectivity and physical hardware used to deliver its services, but also the personnel needed to manage it. After all, what use is a cloud service that provides neither the capacity nor the capability promised by the on-demand, as-a-service nature of cloud?

Hyperscalers don't actually have unlimited capacity – they must plan capacity in advance and provision servers and (to a lesser degree) employees for when they are needed. If it has bought too much capacity, then resources sit wasted in a datacenter – a sunk cost. Too little, and consumers won't get the capability and performance they need, and the cloud provider has missed out on revenue. The profitability of all cloud providers is related to this balancing act – the more accurate a forecast of capacity, the better the margins and the better the customer experience.

What we don't know for certain is what the pandemic will do to overall demand for cloud services. If demand outstrips supply, then users won't just be disappointed – they might be disillusioned and feel that the promise of scalable 'unlimited' capacity is a fallacy, and that, when the chips were down, cloud was not able to deliver. If supply outstrips demand by too much, then margins will be impacted – the provider has servers and engineers sitting unutilized, now with less revenue coming in.

Signs exist that demand is outpacing the average consumption of cloud resources, as those in isolation turn to streaming media, video games, news websites and online retail as a means of escaping boredom, as well as to the web as a means of remote working and selling products and services. Microsoft claimed a 775% increase in calls and meetings in Microsoft Teams in Italy as a result of shelter-in-place orders. Netflix (which uses AWS infrastructure) reduced the bitrate of its streaming video to reduce the load on global connectivity as a result of increased demand. Likewise, Google has similarly reduced default streaming bitrates for similar reasons. According to our COVID-19 flash survey, 34% of enterprises are spending more on IT resources compared with just 3% that are spending less.

However, the question of whether demand will outpace supply is trickier. No matter how intelligent the algorithms or people used to forecast capacity, this is difficult to predict – and new hardware capacity isn't easily found with travel restrictions and crumbling supply chains. Microsoft has stated that European, Asian and South American regions haven't been living up to their 99.99% deployment success target, and users should keep trying if they fail. However, most cloud providers appear to be coping well.

AWS Spot Instances allow consumers to purchase virtual machines at low cost, but with no guarantee of availability. They are cheap because they can be terminated by the provider with little notice. In theory, they are used by cloud providers to sell excess capacity – it's better to fill servers with paying customers, even though customers are paying very little. If another customer wants to pay more than 'the on-demand price,' they have priority over the spot instances. We don't know how AWS's spot pricing algorithm works, but it seems likely that – if the company was struggling with capacity – it would increase the spot price of instances will the aim of reducing the burden on limited capacity. Although pricing does change from time to time, a quick browse through AWS spot history from the past three months has identified no obvious correlation of price increases with COVID-19. In a similar vein, there have been no obvious price drops suggesting an excess of capacity. This implies AWS is handling capacity challenges well.

Hyperscalers Will Survive This

We think price changes in the six months will be unlikely to take place, apart from specific promotions for COVID-19 research or to support struggling businesses. Cutting prices now in a time of uncertainty could be dangerous. If prices are cut, but then there are difficulties with capacity, supply chain or labor, internal costs may rise, leading to reduced margins. If prices are hiked at a later date, then customers that are already embedded with the cloud provider will be penalized. It's better for hyperscalers to keep prices constant for the few months, rather than to cut them now and have to raise them later.

Will cloud providers be impacted by supply chain issues over the long term? It's definitely possible. Servers are created from a multitude of materials and components sourced from all over the world. This supply chain is fragile, and trying to plan capacity under these circumstances is like driving in a hailstorm at night without headlights. But since demand is already very high, we can't envision it spiraling significantly more – after all, most of Europe and the US are trapped, with Asia still only just removing restrictions. If cloud providers are handling capacity today, then we think they've probably got enough until this all blows over.