Back to the edge

In the opening years of this century an online bookseller realised it could make money selling its internally developed virtual server technology to others. Cloud computing was certainly around before Amazon Web Services, but it took off in earnest after the arrival of Amazon’s Elastic Compute Cloud

Before long, established computer industry giants like IBM, Oracle and Microsoft saw their world upended by a rival no one saw coming. They all switched strategies and have since spent billions building their own clouds to catch up with AWS. They have their work cut out. AWS’s revenues are greater than the next four cloud suppliers put together. 

The key idea behind cloud computing is that it offers unprecedented economies of scale. 

Instead of building their own capital-intensive data centres, companies can use someone else’s and share the overheads. They can buy computing power, storage and other digital services in much the same way they buy electricity or water. The resources are there when you turn on the tap. All you need is a credit card to pay the bill each month. 

 

Cloud computing is highly centralised. Where there were once millions of small data centres scattered all over the world, now these servers and applications, which do all the hard work, are in a handful of massive sites close to cheap power and other essential resources. 

In technology, runaway success often sows the seeds of its own destruction. No one is suggesting AWS is in trouble, but it now looks as if there are limits to how far the cloud can go.

Streaming-video giant Netflix is one of AWS’s largest customers. It uses the company’s central cloud to serve up movies and television shows to around 100 million subscribers. But Netflix, like other digital businesses, bumps up against a limitation of AWS. Often the big, efficient regional AWS servers are a long way from customers and that’s a problem. 

Customers naturally want the best possible viewing experience. With the arrival of fast fibre networks, with speeds of up to a gigabit, New Zealand’s internet users expect to see a matching performance improvement from services like Netflix. They reward the companies who deliver this with loyalty and repeat business. They punish those who disappoint by looking elsewhere.  

It turns out that the best way to deliver great online experiences from streaming video or gaming services is to move the point of distribution as close to the customer as possible — right to the edge of the network.    

Enter the next big thing: edge computing. In effect, this means the next season of your favourite television show will be served from somewhere in your suburb or nearby, not from a giant data centre housed on another continent. There are already edge computing data centres in some of Chorus’s exchange buildings. 

Edge computing isn’t just about streaming video, however. The technology is essential for driverless cars. Latency is a vital issue with these. You really don’t want your autonomous car asking a server in Stuttgart for permission to turn left or to apply the emergency brake. Likewise, edge computing will serve drones, robots and all the billions of other Internet-of-Things devices being connected to global networks. 

The cloud will still play a role. Data collected at the edge will continue to be sent to the centre for aggregation and further analysis. But digital intelligence will be more dispersed. 

There has always been ebb and flow in the computer business. The first mainframes were giant central machines. Then came dispersed minicomputers, followed by even more dispersed PCs, phones and tablets. The advent of the cloud saw the centralisation tide come in again, with edge computing it is now heading back out. 

 

Bill Bennett