Enterprise networking is quickly shifting from a hardware/software construct to all-software. This means configuration management, system relationships and other architectural elements will transition from today’s admin/management approach to software development processes in short order.
And that opens up the very real possibility that network oversight, like virtually everything else in data infrastructure, will be subject to the whims of artificial intelligence (AI).
AI is already making significant headway up and down the enterprise data stack, primarily due to the fact that the emerging digital ecosystem driving big data and the Internet of Things (IoT) is simply too complex and too fast-paced for manual control. Increasingly, network resource configuration and consumption will be left to intelligent algorithms embedded into the applications that drive data productivity.
To date, however, very little of this activity has centered on local data center infrastructure. But it is starting to permeate the wide-area networks (WANs) that connect the enterprise, the cloud and the edge.
Recently at the Mobile World Congress in Barcelona, Spain, Aricent announced that it was joining the Central Office Re-architected as a Datacenter (CORD) community to lend its expertise in crafting a reference architecture for AI, hardware abstraction and network programmability. The company says it hopes to empower an entirely new data center based on composable infrastructure, white-box servers and open-source software. The company has taken a keen interest in CORD’s Virtual OLT Hardware Abstraction (VOLTHA) project, and it is contributing a real-time performance monitoring framework that will incorporate intelligent analytics that allow the network to optimize itself continuously for a better user experience.
Meanwhile, Netronome’s Agilio CX SmartNIC is bringing intelligence into the network functions virtualization (NFV) platforms of Juniper, Mirantis, Nuage Networks and others. As telcos become more adept at offering data center network services, Netronome is looking to bolster their ability to carry higher data loads under increasingly complex virtual network architectures. The company says it can boost data center productivity six-fold per dollar spent, while at the same time maintaining vendor independence and complete mobility of virtual machines and virtual network functions. The Agilio platform comes in 10, 25 and 40 GbE versions and provides virtual switch and router datapath offloading for network overlays, load balancing, telemetry and other functions.
The most obvious place to layer AI in a software defined network is the orchestration module, but a start-up called Aria Networks is aiming for the controller instead. The company recently announced a deal with China’s BOCO Inter-Telecom to implement intelligent algorithms, originally developed for genetics research, on BOCO’s cross-domain SDN controller. As explained by SDX Central, this approach places network decision-making on a local footing, which should scale better on the IoT edge. At the moment, the system is tailored for path computation and capacity planning, although additional functions are likely as the algorithms mature.
But AI is not only useful for connectivity. It has profound implications for network security as well. According to IDC’s Jarit Sidhu, an intelligent network will be better able to guard itself against intrusion from the numerous attack vectors that arise from pushing enterprise infrastructure to the edge. The dynamic nature of these evolving device-centric networks will require a constantly shifting security posture and deep packet analysis to identify and isolate malicious code, and perhaps even track it back to its source. Intelligent, even cognitive, technology with its ability to adapt and respond to changing conditions on its own, offers the only real chance of managing such complexity in an intuitive manner.
In many ways, artificial intelligence is a lot like virtualization. It can be useful in limited deployments, but it really needs to permeate across data infrastructure to produce truly significant benefits.
With an intelligent network, the enterprise will not only be able to keep up with the pace of a digital economy, it will steadily increase its own unique value proposition to the world at large. So even if someone else is larger or faster, they will never be able to completely match what the intelligent enterprise has created for itself.
Arthur Cole is a freelance journalist with more than 25 years’ experience covering enterprise IT, telecommunications and other high-tech industries.