Making predictions is a risky business, especially in tech, which changes rapidly and unpredictably. Even more so at the edge, which is so new and still emerging. But predictions about the edge are also more valuable. If we can better understand what’s coming, we can be better prepared, says Khalid Raza, CEO and founder of Graphiant.
Data Gets Stranded at the Edge
We all know that data is exploding, especially unstructured (or file) data. IDC predicts we’ll have 175 zettabytes by 2025. That’s not news, but there is an edge angle to this statistic. It has to do with the fact that bandwidth is growing more slowly than data. And that is important at the edge.
If we took all the data in the world (estimated as 113ZB today) and transferred it across the street today, using all the bandwidth in the world (estimated to be 786 Tbps today), it would take 36 years to complete that transfer.
But flash forward a few years – to 2025, that same transfer (all the data in 2025 – 175ZB using all the bandwidth in 2025 – 1.687 Pbps) would take 70 years.
Why does this matter at the edge? Because we can no longer centralize data. Instead, we have to start storing it where it is generated and used. Since most data is generated by IoT devices sitting in factories and other edge locations, this means storing data at the edge.
This also means organizations will start creating data lakes at the edge, and building AI models at the edge, to keep from having to backhaul all that data up to the cloud or data center.
Networking Technologies Converge
The migration from the traditional hub and spokes networks to today’s any-to-any networks has caused enterprises to add many layers to our connectivity tech stack. Currently, enterprises need at least four solutions or network constructs to provide service to their users:
- MPLS and/or SD-WAN to connect locations
- SSE for cloud-delivered security (SWG, DLP, etc.)
- SDIC for cloud connectivity
- DMZs for B2B connectivity
This is costly and inefficient. The good news is that in 2023, we’ll start to see these separate network constructs converge, and enterprises will have the option to consume a single service or multiple services. Either a single vendor SASE or a partnership between SSE and NaaS providers will enable this functionality.
The Emergence of Network-as-a-Service
Before I explain this one, it’s important to note that the Internet has come full circle. In 1962, J. R. Licklider – working at DARPA – wrote a position piece titled “The Intergalactic Network.” This title was ambitious, given that nobody had yet succeeded at networking even two computers together. But Licklider was prescient. His work predicted digital libraries, e-commerce, online banking and cloud computing.
Originally Licklider conceived this Intergalactic Network (which became ARPAnet at first, and then eventually became the Internet). It was very much a peer-to-peer network. The reason was simple – he wanted the network to survive in case of a nuclear attack.
Fast-forward 60 years and consumer applications (such as Facebook and Amazon) have remolded, causing the Internet into a client-server model. Compute and data were concentrated on massive servers, and clients were distributed around the globe.
Call this the consumer Internet. You can see evidence of this in Internet usage statistics:
- Nearly half (49%) of Internet traffic today is video and music sharing.
- Social media is next, at 19%.
- Web is 13%.
- The rest is split between messaging, search, gaming, and – in a small way – business computing.
- Note that mobile devices comprise HALF of the Internet traffic and consumers downloaded 204 billion mobile apps in 2019 alone.
But we are beginning to see a shift to the industrial Internet where – as previously discussed – the center of gravity has shifted to the edge. At the same time, we’re moving to a peer-to-peer model. That is the kind of network enterprises build! There are massively higher numbers of nodes, and solving for performance, reliability, security and privacy is excruciatingly difficult.
The Rise of On-demand
Instead of designing, purchasing, building and maintaining bespoke networks, enterprises will configure the network on a cloud portal and consume what they need on demand. The question isn’t why this will happen but rather what took so long.
We already consume cloud resources, storage, apps and services this way. Why not networks? This will allow the Enterprise to concentrate on policies and the as-a-Service network vendor to focus on building and delivering a “hyperscaler-class” network.
Benefits to look forward to
This will vastly improve things.
First, enterprises won’t need to have as much networking expertise in-house. Network architects will focus on setting business objectives and leave network-building to the experts at the as-a-Service network architects.
Second, because this new industrial network will be private (a la MPLS networks), predictable, reliable performance will be guaranteed. Enterprises can even get network performance SLAs, which overlay networks like SD-WAN cannot provide.
Third, security and privacy are assured because data will always be encrypted in this new network.
And finally, this new industrial network will become much more affordable than MPLS, and because network architects no longer need to build point-to-point tunnels, it will be easier and faster to provision.
The first two predictions either force this change (as in the case of data growth forcing a move to peer-to-peer networking) or are enabled by the “as-a-Service” networking model (networking technologies converging). 2023 is going to be a very interesting year!
Image Source: Shutterstock