Five years ago, an infamous Gartner blog said, “The Edge will Eat the Cloud.” The gist was that while everyone was rushing to the cloud, Gartner saw a rush in the other direction – to the edge–driven by a need for lower latency and near-real-time processing. Fast forward five years, and with edge computing now forecasted to be an $800B market by 2028, it’s a perfect time to revisit the topic.
So, has there been a rush to the edge? Absolutely. Is it “eating the cloud?” Well, the cloud is as strong as ever, so it is hard to make that case. But there is a strong chance the edge will transform the very internet itself. To explain why, let’s go back to the beginning of the internet.
The intergalactic computer network
On October 4, 1957, the USSR launched the world’s first satellite, Sputnik. Four months later, a shocked US Department of Defense established the Advanced Research Projects Agency (ARPA) to ensure the US never lost another technology race. While landing on the moon may be the most famous outcome from ARPA, the most impactful outcome was the internet.
This began with a research paper written by J. R. Licklider in 1962. It foresaw so much of what would eventually become the Internet, including e-commerce, online banking and cloud computing. The first iteration was called ARPAnet (for obvious reasons). ARPAnet was conceived as a peer-to-peer network, designed with no central core so that it could withstand a nuclear attack.
At first, ARPAnet was restricted to use by the government and educational institutions. But in 1991, ARPAnet became the internet and was eventually made available to the public. At first, the internet was still a peer-to-peer network. But over the next three decades, the internet increasingly became a client-server architecture used mostly for consumer applications.
A consumer-grade network
Today’s internet usage underlines its conversion to a consumer network:
- Nearly half (49%) of internet traffic is video and music sharing.
- Social media is next, at 19%.
- Web usage is at 13%.
The rest of the traffic is split between messaging, search, gaming and — in a small way — business computing. Note that mobile devices comprise HALF of all internet traffic and consumers downloaded 204 billion mobile apps in 2019 alone.
What does a consumer look like? Overwhelmingly client-server. Nearly all consumer applications (Facebook, YouTube and Spotify) are client-server. Data and compute are co-located at the internet’s core, while users are at the edge.
Client-server is fine for a consumer-grade network but it has characteristics that, while benign for consumers, are deadly for business applications:
- The movement of data is subject to the laws of physics. The longer it has to travel, the higher the latency. The endless routing in today’s client-server model exacerbates this.
- Massive client-server architectures are highly susceptible to random outages and brownouts in the “digital wilderness” that sits between the client (e.g., the consumer) and the server (e.g., Facebook).
- Security and Privacy Concerns. The threat surface in a client-server consumer internet is massive. Private, sensitive data is in motion for huge amounts of time over vast distances, exposing this data to security and privacy threats.
- Data Mobility Chasm. There is a subtle problem growing that is just starting to be understood. It is becoming impossible to move data around the internet quickly. The problem is that data is growing faster than bandwidth. If we apply all the world’s bandwidth to transfer all the world’s data across the street, it will take 36 years to complete. And it’s getting worse! By 2025 it will take 70 years!
The new business-grade network
Consumers can live with these challenges, but businesses cannot. In fact, the rise of edge computing is driven – in large part – by these very challenges (latency, security, data mobility issues). Sixty years after J. R. Licklider imagined the internet, we are watching a reimagining. Flipping compute to the edge changes so much:
- The enabling of ultra-low latency apps to run (e.g., remote robotic surgery).
- A reduction in the threat of outages and brownouts because traffic no longer traverses the massive digital wilderness of the public internet.
- A drastically reduced threat surface, thereby improving data security and privacy.
- The edge exempts businesses from the gridlock of the data mobility chasm — data can stay local and doesn’t need to migrate around the internet.
Many new technologies are helping to build the edge. 5G and Wi-Fi 6, for example, both promise speed and low latency. But there is still one piece we’re missing. The speed and low-latency advantages evaporate if used in a client-server consumer network. To fully realize the promise of edge computing – and to build a true business internet, we also need to move to a peer-to-peer edge computing model to build a true business internet.
What changes in the new business-grade network?
So, how does this new business-grade network have to operate? Four things need to change:
- Because there are so many more connections we need to provide in a peer-to-peer network, we need to make it much easier to create these connections. Instead of building a bespoke network that requires routing, tunnels, and traffic engineering for each connection, we need to provision networks as-a-Service the same way we provision other resources like compute and storage.
- We need a private network – like MPLS – that handles routing for us and doesn’t require encryption because it isn’t public. This will also provide predictable performance, allowing for the enterprise-grade SLAs we need.
- We must make connecting with partners and customers simpler – without complex and static DMZs that need complex security configurations.
- We need an easy way to provision cloud connectivity. This connectivity must make the cloud a point in the network, not the “server” of our “client-server” network.
The edge is eating the internet
So, was Gartner right? Is the edge eating the cloud? No, but the edge is eating the internet. We’re seeing a transformation from a consumer-grade client-server internet into a business-grade, peer-to-peer network. A fast, low-latency, reliable, secure network. The network Licklider first imagined six decades ago.