Continuing my previous blog post in Part 1, Self-Driving Network is a network automation concept invented by Kireeti Kompella of Juniper Networks. Kompella first launched this concept in Berlin, Germany in July 2016. I happened to sit in the front row listening to his interesting presentation, so I have first-hand knowledge of what he means by a Self-Driving Network.
In general, a Self-Driving Network would:
- Accept “guidance” from a network operator
- Self-discover its constituent parts
- Self-monitor using probes and other techniques
- Auto-detect when a new service is needed and auto-enable it
- Automatically monitor and update services to optimize service delivery
- Use machine learning for introspection (self-analysis)
- Self-report periodically or when an unexpected situation arises
Here is the motivation behind it. Network complexity is increasing exponentially as traffic levels continue to grow and new devices proliferate. This manifests itself as rising operational costs and slower time to revenue, squeezing margins for traditional service providers. Abstracting, simplifying, and obscuring this complexity are major challenges for the industry going forward. While the industry has been making modest, incremental progress in the areas of automation and virtualization, we want disruptive innovation. Automation is a great first step, but we need much more.
There is an economic imperative for traditional service providers to radically change the way they operate their networks. Disrupt yourself or, eventually, others will.
The primary cost for traditional telecommunications service providers lies in their operating expenses, not capital expenditures. The annual cost to operate a network is much higher than the cost to initially buy the equipment. As networks become bigger, more complex, and tougher to manage, these operating costs skyrocket.
To address the increasingly difficult network economics, first we need to more aggressively adopt automation techniques and then push toward networks that operate autonomously. The benefits of Self-Driving Network are threefold:
- Reduce operational complexity by simplifying and abstracting the network.
- Enable customers to deploy new network services faster.
- Improve capacity utilization and network resiliency through deep telemetry.
Self-Driving Network Automation Levels
Juniper Networks classifies different levels of network automation, as follows:
- Level 0 – No Automation (“Manual”): Automated system issues warnings and may momentarily intervene but has no sustained network control.
- Level 1 – Automated Network (“Hands On”): Operator and automated system shares control over the network. The operator must be ready to retake full control at any time.
- Level 2 – Fine Grained Monitorization (“Hands Off”): The automated system takes full control of the network (configuration, monitoring, and troubleshooting). The operator must monitor the network and be prepared to immediately intervene at any time if the automated system fails to respond properly.
- Level 3 – Analytics (“Eyes Off”): The operator can safely turn their attention away from the management tasks, e.g. the operator can sip a cup of coffee while reading this article. The network will handle situations that call for an immediate response, like emergency repair of the network. The operator must still be prepared to intervene within some limited time when called upon by the network to do so.
- Level 4 – Autonomous Process (”Mind Off”): As level 3, but no operator attention is ever required for continuous operation of the network. Self-healing is supported only in limited situations or under special circumstances, like network congestions.
- Level 5 – Self-Driving Network (”Wheel Off”): No human intervention is required on any of the controls, configuration, monitoring, troubleshooting, self-healing, or else.
These different levels of classification are illustrated in the following table. Note what happens in the shift from SDN 2 to from SDN 3: the human operator no longer has to monitor the network. This is the final aspect of the ”network management task” that is now passed over from the human to the automated system. At SDN 3, the human operator still has the responsibility to intervene when asked to do so by the automated system. At SDN 4 the human operator is relieved of that responsibility and at SDN 5 the automated system will never need to ask for an intervention.
After you understand the different levels of network automation, the next question is how do you get your network to reach the ultimate automation level, the Self-Driving Network?
Roads to Self-Driving Network
As with self-driving cars, Self-Driving Network does not happen overnight. Automobile evolves slowly from their primitive forms in the 1900s until a much more sophisticated function nowadays. The key here is incremental and continuous improvement, which the Japanese call Kaizen.
Similarly, we need to evolve our network gradually from its state now towards the Self-Driving Network in the future. The key here is incremental and continuous improvement to the network by adding more automation features and functionalities over time towards higher levels of automation.
Since there are 5 levels of network automation, there are also 5 steps of Kaizen towards Self-Driving Network, as follows:
- Level 0 – Level 1 Automated Network: If you start from a fully manual network, you start adding automation tools and frameworks to your network, such as NETCONF/YANG, OpenConfig, Zero Touch Provisioning, RESTful API, to reach Level 1.
- Level 1 – Level 2 Fine Grained Monitorization: You then add fine grained telemetry capabilities such as Juniper Telemetry Interface or OpenConfig and their related telemetry collectors to reach Level 2.
- Level 2 – Level 3 Analytics: Supplementing telemetry with network health monitoring, capacity planning, and resource monitoring solutions will give you analytics to reach Level 3.
- Level 3 – Level 4 Autonomous Process:Adding Event Driven Infrastructure (EDI), automatic service placement, and self-healing abilities to your network will bring it to Level 4.
- Level 4 – Level 5 Self-Driving Network: Finally, complementing all automation processes with a closed-loop intent based networking (IBN) orchestrator, such as the Juniper Elastic Edge (E2) Controller, will complete the network automation transformation to Level 5.
The evolution towards Self-Driving Network is illustrated in the following diagram.
Self-Driving Cars Need Self-Driving Network
Before we finish, I would like to draw another direct correlation between self-driving cars and Self-Driving Network. Self-driving cars require constant low-latency communication to the network transmitting enormous real-time latency-sensitive data.
In other words, self-driving cars are completely dependent on the network, in that a small disruption in the network communication could result in unintended consequences of the self-driving cars operation and safety.
Since self-driving cars are automated, they need networks that are automated, too. The conclusion is self-driving cars are not possible without robust networks; self-driving cars need Self-Driving Network because:
- Networks are required within cars to connect local telemetry to local intelligence and control.
- Networks are required between cars on the road for safety and amplified intelligence, which comes with the interconnectedness.
- The macro network to the cloud is required for centralized intelligence, analytics, management, and control.
There you have it, folks! As promised, we have an easy and straight forward explanation of the different automation levels of a Self-Driving Network. Now that you also fully understand the different levels of self-driving cars, you will know what to ask when buying a self-driving car in the future from Tesla and the likes. You may also want to ask if your future self-driving cars – that you won’t be driving – will be connected to a Self-Driving Network.
Good luck on your automation journey, network, cars, and all that!