The impact of global network trends on WAN services

When you look at the expansion of major cloud providers over the last 15 years - AWS, Facebook, Google, and Microsoft - you’ll notice an amassing of data centers in the US, Western Europe, and parts of East Asia. But until recently, there’s been barely anything in the global South.

This is largely because much of the early growth in cloud data centers was geared towards developed markets and hubs. AWS launched its first cloud region on the east coast of the US in 2006, and was quickly followed by Microsoft in 2008, and so the bulk of global cloud infrastructure was based in North America and Europe.

In the years that followed up to 2015, around 3 to 4 new data centers were launched per year. That quickly accelerated between 2015 and 2018 as 15 data centers a year were built largely outside the established hubs. In 2019, fuelled by Oracle Cloud, 28 new cloud regions were launched. Today, what we’re seeing, as a result, is the extraordinary growth of global networks across Africa, South America, and other regions that were previously left behind.

So how have we got here? And what does it mean for businesses looking to connect to cloud services in the future? Let’s take a look at how this boom in connectivity has fundamentally altered the future of the cloud.


Shifts in inter-regional capacity

To understand the impact of cloud geography on the WAN, it is worth looking at shifts in inter-regional capacity between cloud regions. Broadly speaking, the biggest shift in global cloud network share over the last decade has been a move away from the dominance of internet backbone providers in favor of the major content providers.

When we’re discussing content providers, there are two types of networks to distinguish between. Inter-data center demand networks, which connect data centers to other data centers, tend to be higher capacity routes with lower prices.

What we’ve seen is a surge in bandwidth on these traditional routes as all the major content providers link up. Many of these are close to dense metropolitan areas, but some are routed outside these regions, often owing to factors such as the regulatory environment.

The second type of network is content or cloud distribution networks, which connect data centers to the end-users. These are typically smaller in bandwidth than the inter-DC routes, and their capacity is geared towards content distribution to end users in both the major and growing hubs.

As you might expect, the Transatlantic route between North America and Europe has the most capacity connecting the two regions, followed by North America and Asia. But what we’re also starting to see is a rapid growth in inter-regional capacity in the Middle East.


Cloud geography and the impact on architecture

So how does all this affect how you connect with cloud service providers, set up your sites, or choose cloud regions to link them to?

When people think of the cloud, they often think of it as ubiquitous. This may be true if you’re located in the US, Europe, or parts of East Asia. If you’re based in South America, Africa, and until recently, the Middle East, it’s a different story.

There are far fewer data centers and cloud regions in these areas. So if you’re trying to connect your sites to the cloud in South America or Africa, you don’t have much choice - you’re going to have to connect to São Paulo, Johannesburg, or Cape Town. The location of content provider data centers has a major impact on the quality and latency of cloud services in these regions.

A lot of the sites that are in these regions are still having to connect to the data regions in the US and Europe, which have been the historical hubs of content for the internet. If you’re outside any of these regions, and you have a data center or two pop up in your area, you might see traffic start to split between regions. So if you’re based in Lagos, you’ll probably see part of your traffic going up to Europe as it is quite a distance south to hit data centers in South Africa, depending on the cloud services available.

One region that has seen a boom in availability zones is Asia - and this is largely thanks to Alibaba’s cloud regions within China. Alibaba has over 50 availability zones in China alone. To put that in perspective, AWS and Google average 20 zones across six or seven cloud regions in North America. However, with Alibaba so focused on Chinese enterprises, their presence is very limited elsewhere in Asia.


The impact of submarine cable investments on WAN access

One of the greatest limiting factors for connecting to cloud apps, at least in the global South, is an extremely limited number of on-ramps in the region. However, this is changing quickly. We’re seeing more and more investments by content providers into submarine cables, such as Google, whose Equiano subsea cable connects Africa with Europe via Portugal, Nigeria, and South Africa.

As end users increase in these regions and content providers begin to see a business case, subsea cables, on-ramps, and caches start appearing - which is a strong indicator that cloud data centers are probably going to be built further down the line. Before these appear, businesses can expect lower bandwidth and a greater focus from content providers on distributing their content.

You can see this in the submarine cable latency that exists today between different regions. While the US - UK subsea cable has an average latency of 58-66ms, South Africa to the UK averages a whopping 140-204ms latency. At that point, the performance of many cloud services becomes noticeably impacted and could cause problems when it comes to accessing cloud apps and services.

It’s worth mentioning that distance isn’t the only thing that matters. For example, even though many cloud regions in Asia are much further in distance than Europe from LA, their latency is much lower or about the same as European locations. Distance doesn’t really explain this discrepancy.

However, if you look at the networks that are connecting the west coast to Asia, they’re hitting subsea cables, whereas if you’re connecting to Europe from LA, you have to go through the terrestrial networks of the US before you hit the transatlantic cable. Therefore, data center proximity to subsea cables can have a real effect on performance and cost.


Mitigating cloud performance issues

Ultimately, cloud service providers pass on network costs to their customers. So if you’re operating in developed markets, you can expect a lot of choice, performance, and lower costs. But if you’re in emerging markets where there are fewer data centers and overall smaller ecosystems, these are the areas where you’re going to see higher costs and performance issues.

To address these issues, you don’t have to wait five years for cloud providers to finish new subsea cables and on-ramps, however. Expereo solutions can improve performance between your sites and cloud providers, ensuring your data goes through the most direct route possible to reach its destination - meaning you can receive cloud performance that matches on-premises. 

Talk to a Cloud expert



These articles might interest you as well