The future is here: The autonomous network

The network is becoming increasingly software-defined and hardware agnostic; at the same time, analytics, artificial intelligence (AI) and machine learning (ML)  are being embedded into the IT infrastructure stack. The convergence of these two developments is paving the way towards the autonomous network - a network environment which requires minimal manual intervention to configure, monitor and maintain.

Networks today are far more complex than they used to be just a decade ago before the advent of devices and sensors. This means that they are plugged into a multitude of sources and destinations for data - from large data centres to drones and AI-enabled smart devices.

This complexity means that human-administered fault detection, traffic rerouting, outage prevention, and other management tasks are no longer the ideal scenarios for managing the network. Additionally, the fact that networks draw so much of their data from intelligent devices and platforms means they have the wherewithal to self-correct themselves, silently making decisions that optimise paths and service errors.

Because of the large variety of proprietary systems that the contemporary network interfaces with, open APIs are an essential component in ensuring that the systems are able to rapidly exchange data and “talk” to each other.

A key element in enabling the autonomous networks is also the ability to process and draw insight from large amounts of data. The entire IT infrastructure stack generates enormous amounts of data that can be pooled via open APIs and fed to AI and ML algorithms. The latest libraries and frameworks that can manage big data at scale such as Hadoop, and its accompanying technologies such as MapR and Spark, can serve an important role in this process.

Another important piece of the puzzle is the cloud: assuming (near) limitless storage and processing capacity enabled by the cloud, cloud-based AI and ML algorithms can paint a holistic picture of the state of the network, which can “learn” as it completes routine tasks and detects aberrant or inefficient processes. Bandwidth is also a critical component - it needs to be provisioned according to demand required by the receiving/sending application or information being transmitted.

Sensors are the “eyes” and “ears” of the network, the sources of data it takes input from and delivers direction to, and constitute a necessary aspect of the autonomous network.

No innovation stands apart from its ultimate value to the line of business, and will not be adopted unless it is seen to provide tangible benefits. The good news is that autonomous networks can greatly increase cost efficiency in this regard, and save many man-hours spent by network administrators in keeping a network operating like a well-oiled machine.

Autonomous networks have a direct impact on business operations, reducing the time required to correct errors and disentangle overloaded network paths and nodes. There are also the indirect benefits provided by all the components needed to enable autonomous networks in the first place - open APIs, AI and ML algorithms, cloud, sensors, and big data analytics.

Especially in Asia’s emerging economies, the adoption of autonomous networking can tremendously benefit the enterprise, as the exponential growth in customer base means ever-increasing network resources, bandwidth and capacity need to be utilised optimally.

Scalability of networks is enhanced by letting autonomous decision-making take a leading role, as manual regulation of routine maintenance tasks slows down the ability for the enterprise networks to grow at a rapid pace. Traffic patterns, especially in fluctuating network capacities and use-case scenarios, can be difficult to manually keep tabs on - a problem better managed by automated analysis and correction if required.

The end user benefits are clear: better response rate and lower turnaround time in terms of support case resolution, smoother transition to the IoT-enabled and 5G future (because of the natural adaptability and scalability of autonomous networks) and much greater intelligence about the state of their systems and endpoints thanks to the precise data generated and transmitted by the “intelligent” self-servicing network.

Finally, the autonomous network is a significant step towards the holy grail of computing: “real-time” IT in every sense, where systems self-correct, self-enhance and grow while human intervention stays largely out of the loop. Because of AI and ML enabled applications, the wide use of sensors and largely virtualised data centers that provision computing and storage at will, this future will become a reality. Networking needs to stay abreast of these developments and avoid becoming the weakest link in the autonomous, real-time chain taking over the IT stack.