Info Image

Shaping the Future of Networks and AI

Shaping the Future of Networks and AI Image Credit: peshkov/BigStockPhoto.com

It’s prediction time again. You’re probably tired of AI predictions by now, so I’ll give you a break – at least for the first 75% of this article. The first three of my four predictions are not AI-related – they are about how networks will change in 2024. Then, I slip in a final prediction about AI. But, to be fair, it is about how AI will change networks.

#1: The Internet Returns to its Peer-to-Peer Roots

The internet was born in 1969 as “ARPANET,” a government-driven project. ARPANET was a peer-to-peer network – anything could connect to anything. There was no concept of servers. But, when the government opened this technology to the public as the Internet, consumer apps and traffic quickly drove the internet to become a client-server network. We have a relatively small number of mega-servers (Facebook, Amazon, Netflix) serving billions of consumers.

Business made do with this architecture because they could. Companies had few locations to connect (the data center, a few offices, perhaps a factory). But business has morphed – there has been an explosion in the number of sites business needs to connect:

  • Remote workers
  • Edge locations
  • Partners
  • Customers
  • Multiple clouds

Of course, this requires a new kind of network that makes it easy to connect anything to anything – no complex firewalls, DMZ, etc. We need a simple way for anything to connect to the central network (the middle mile) and then – from there – to connect to other nodes on the network.

And that leads to my second prediction …  

#2: The RISE of the Network-as-a-Service (NaaS)

The public cloud (AWS, Azure, et al.) has revolutionized storage and compute for businesses. In 2024, NaaS will do the same for networking. The drivers of this trend are many:

  • The number of things business networks need to connect has exploded (more sites, remote workers, edge locations, partners, customers, multiple clouds). See my first prediction for how peer-to-peer is the new normal.
  • Business transformation requires a new connection to happen much more quickly – in minutes, not weeks.
  • Businesses have become accustomed to the no-CapEx pay-as-you-go economics of cloud.
  • A faster pace of business requires a much more elastic network (able to scale up and down quickly as business conditions change).

A Network-as-a-Service (NaaS) addresses these challenges. But it is a big change – are enterprise IT leaders ready? A 2023 survey says yes – 7 of 8 say it is extremely likely they will switch to a NaaS.

Further, a true NaaS makes sense for service providers as well:

  • Allows business expansion with very low capital (new geographies, new services, etc.)
  • Moving to NaaS frees rack space for compute to differentiate the provider’s services

But what about that old stalwart, MPLS? Are we finally saying goodbye to MPLS? Well, not yet as it turns out. MPLS is the subject of my third prediction...  

#3: In 2024, MPLS Switches from Middle Mile to Last Mile Solutions

MPLS has traditionally been a highly scalable middle-mile solution. However, as we look ahead to 2024, Network as a Service (NaaS) solutions are set to reshape the landscape. The reasons behind this evolution are evident. When transitioning to a robust public NaaS, the distinction between MPLS and SDWAN becomes less relevant. NaaS, built on a foundation of global connectivity, offers the necessary flexibility. If the last mile connection is public, it becomes SDWAN, and if it's private, it becomes MPLS. The choice between the two will be determined by the specific application requirements, rather than any technological limitations.

Therefore, the frequently heard statement that MPLS is dead is not accurate. MPLS will adapt by incorporating SDN principles, particularly state abstraction. By eliminating customer profiles from every middle-mile router, unlike traditional PE (Provider Edge) routers, we will transition towards a true NaaS solution in the middle mile. In 2024 and beyond, MPLS is poised to remain a highly effective and scalable solution.

And that AI prediction I promised you? 

#4: In 2024, AI will Give Your Network a Voice

AI represents the most substantial transformation the internet has undergone since its inception. I view AI in the context of networking through two distinct categories: AI for Networking and Networking for AI.

  • AI for Networking involves leveraging AI to support our existing networks. AI algorithms can analyze vast amounts of real-time network performance and operational data. This analysis can be instrumental in network planning, optimization, and efficiency. AI can also offer predictive maintenance, allowing us to address issues before they disrupt network operations. One significant concern arises from the proliferation of distributed data generation. AI plays a crucial role in securely processing and transporting this data.
  • Networking for AI focuses on building networks capable of meeting the security, privacy, and scalability requirements of AI workloads. Software-Defined Networking (SDN) has given us the ability to abstract network states. Building a network without customer profiles in transit, running on commodity hardware, and capable of handling data exchange between entities is imperative. Customers will formulate their policies concerning distributed data location and securely transport data to their private cloud or exchange it with their business partners in the middle mile, creating a more private and secure data exchange environment.

As we move forward into the next generation of infrastructure, there's a growing realization of the power-saving potential that SDN was intended to offer. In the distributed middle mile, which is poised to become a significant AI data processing hub, deploying data plane boxes without the need for Virtual Network Functions (VNF) per customer can result in substantial power and real estate savings. Rather than relying on large networking routers or customer network VNFs, the middle mile can facilitate B2B data exchanges for AI workloads by prioritizing workload-centric compute and storage solutions, reducing power consumption and real estate requirements need to just run network infrastructure

So, there you have it. 2024 is a brave new world for networking, but it is all for the better. We’ll talk again in 12 months when I see how these predictions fared and we look to 2025.  

NEW REPORT:
Next-Gen DPI for ZTNA: Advanced Traffic Detection for Real-Time Identity and Context Awareness
Author

Khalid Raza, CEO and Founder of Graphiant, was the co-founder and CTO of Viptela and is widely regarded as the “Father” of SD-WAN. He is a visionary in routing protocols and large-scale network architectures. In the past 25 years, he has redefined expectations and delivered innovative solutions across all industries for the Tier-1 carriers and Fortune 100 companies.

PREVIOUS POST

Push to Eliminate 'Digital Poverty' to Drive Demand for Satellite-Powered Broadband Connectivity Post Pandemic