You might have heard the term ‘edge computing’ cropping more and more in advertising circles in recent times, but the concept isn’t a new one.
The basic principle, of moving the processing of data away from massive data centres and closer to where it’s physically collected, emerged in the 90s with the development of content delivery networks or CDNs – which play a major role in video delivery.
But technological changes are driving higher usage of edge computing. In 2018 Gartner found that just ten percent of enterprise-generated data was processed through edge computing, but Gartner predicted that this would rise to 75 percent by 2025.
And the advertising industry is being caught in this wave, with edge computing set to play a bigger role in a number of areas across the industry.
The Basics
Historically, companies which process huge volumes of data would use massive centralised data centres.
But these data centres may be thousands of miles away from the end user. Even though data can be delivered over an internet connection extremely quickly, these data transfers still cause latency issues.
And when massive amounts of data are being sent at the same time, bandwidth limits can be an issue.
The concept of edge computing means reducing the distance data has to travel, and the volume of data which needs to be sent, by processing as much of this data as possible closer to the source of the data.
This model of data processing can speed up decisioning for data-driven advertising, speeding up load times for users. This will become more and more relevant as more and more data is fed into the ad decisioning process, as connected home devices play a bigger role in advertising.
And edge computing is also being used to improve privacy in online advertising, reducing the amount of personal data which is sent away from a user’s device.
The Technical Details
Edge computing architecture can take several different forms.
In some cases this means building ‘edge nodes’ (a local server and data centre) which are closer to the end user. These edge nodes will then process as much of the data as they can themselves, and only send data to the cloud or to a centralised data centre where necessary.
This is the principle behind CDNs. By having a distributed network of servers and data centres, internet-connected devices communicate with whichever node is closest, rather than all having to communicate with the same data centre which might be half the world away.
In other cases, data might be processed on the user’s device itself. This is becoming increasingly common as the processing power available on consumer devices continues to grow.
For example we’re seeing this happen on smart home devices, which ingest large amounts of complex data, and are expected to have quick response times. At the moment, a smart home device like an Amazon Echo will send data off to the cloud to process a users’ request and get back an answer.
But Amazon is working to process more data on its Echo devices, thus reducing the amount of data being sent and speeding up response times.
The first of these two models already plays a big role in advertising. As mentioned, CDNs are used to deliver video content, including video ads.
But it’s really the latter model, where more data is processed on a consumer’s devices, which is generating excitement in the advertising world.
We’re already seeing edge computing being used to enhance user privacy. If a user’s device is able to use personal data for ad decisioning, but without that personal data ever leaving their device, third parties can use that data without actually being able to see it.
Apple for example staked its claim to the privacy high ground partly by keeping a lot of personal data on users’ devices, using on-device technology to interpret and process that data. And some of Google’s privacy sandbox proposals preserve privacy by handling ad decisioning on a user’s device.
And edge computing is expected to play a growing role as internet-connected devices proliferate.
As connected-home devices become more common, they’ll collect huge volumes of data on their owners. Not all of this will necessarily be relevant to advertisers (it’s unlikely that many brands will want to specifically target ads to users who burnt their toast in their connected toaster that morning). But a lot of it is likely to be of some interest to marketers. Edge computing will make it much quicker to process this data, and potentially avert privacy concerns by keeping sensitive data within a local network.
The Pros and Cons
The main benefits of edge computing have already been outlined above: its speed, lower use of bandwidth, and privacy protection.
But there is a reason that edge computing hasn’t been the default architecture for data processing.
The main disadvantages are the extra burdens placed on consumers’ devices, since they have to be able to store and process data themselves. These limits can mean the sophistication of this data processing, and the amount of data which is stored, could be reduced compared to systems where data is processed off-device.
