Our networks have matured over the years, and as we roll into a new realm of 5G, we’re left revisiting, revising and even reinventing our approach to service level agreements (SLA) and assurance. Bluntly put, our current approach to SLA assurance is, well, not very assuring. We’re accustomed to just meeting agreements with a best-effort approach by over-provisioning the network with more bandwidth than necessary, to ensure that there is capacity when needed.
But in today’s agile world where advances in computing, artificial intelligence and machine learning have dramatically changed how we do businesses, our perspective network management and assurance needs to move beyond a static, hands-on approach to a more agile, intelligent and self-organizing way to manage SLA assurance.
The 5G pressure cooker and the need for change
Anyone in networking knows that 5G isn’t just the next G. It means having to rethink our approach across the entire wireless (and fixed) ecosystem. The massive shift in communications promises to be the first network of its kind to unite networks, devices and applications like never before.
As we move closer to the age of 5G, it’s likely we’ll see service requests change on a number of levels. Not only will they become increasingly sophisticated — through a large variety of edge devices – but they will come from human customers and machines alike, for a large number of business and personal applications.
Take, for instance, how 5G networks will need to quickly, reliably and automatically accommodate fluctuations of high bandwidth service requests like remote surgery and remote operations in the field, with the day-to-day requests from businesses leveraging 5G for seamless video conferences or consumers streaming multiple feeds within their home.
We’ll also see devices, users and services each with unique profiles for predefined service levels that correspond to different levels of network capabilities (SLAs). Here, a user can select the relevant profile based on the service they need. In practice, these can change each time a service is selected, and will no doubt change regularly, as machines and automated applications increasingly handle these service requests.
In this new ecosystem, setting the pre-defined patterns for service demand, as well as for the traffic that services generate, becomes simply unfeasible. The situation becomes even more complicated when we consider the rapidly expanding IoT ecosystem – predicted to grow to more than 20 billion devices in 2020 — nearly doubling the number of connected devices in use today.
Why 5G thrives on dynamic and assured
How do you plan and commission a network that not only supports dynamic and assured connectivity services, but also has the ability to provision services itself? If we were to build a network like this today, with existing methods, it would likely fail. Not only is resource use usually inefficient, but we are often unable to meet even simplest of user requested changes. Today’s services are mostly binary – on or off. And when on, best effort comes to play.
5G networks, on the other hand, will be required to meet requests for dynamic, assured connectivity services – it’s integral to the promise of 5G. It’s time to break down, however, dynamic and assured.
Dynamic: A flexible network with the ability to configure and provision in real-time is critical to 5G. Communication service providers (CSPs) need a dynamic, responsive network to be able to gain a competitive edge, especially against cloud-based and over-the-top (OTT) providers who already provide services dynamically.
Assured: Assurance is the ability of the network to guarantee services are performing to the agreed SLAs, especially in the areas of bandwidth, latency, and availability. In the age of 5G, best effort doesn’t make the cut, as it’s no longer sufficient just to recognize when levels aren’t being met and reimburse the customer. In fact, moving beyond best effort is a major defining component of many 5G services, demonstrating the need for the underlying network to be self-monitoring and self-healing. Because this strays into the realm of intelligent networking (requiring control of the underlying network), OTT providers cannot provide this.
A network that does its own work?
Beyond 5G, there’s no bigger hype than artificial intelligence (AI), and this is bleeding over into telecommunications. Leveraging AI for networking promises a positive upside to self-provisioning, self-healing networks. However, there is plenty of confusion between what is possible and what is still out of reach. Achieving a dynamic and smart network isn’t as simple as adding machine learning by placing a deep neural network (DNN) on today’s antiquated, static and mostly manual networks. A fully operational and truly AI-based, agile and self-organizing network will require a whole lot more.
A network that is truly assured and dynamic needs to be agile and self-organizing – made up of both a programmable transport network and agile control layer. The result is a network of the future, ready for 5G services, where human intervention is kept to a bare minimum:
The transport network: This foundational component continues to be responsible for providing “physical” service connectivity. For 5G, this entails transporting countless streams of information-carrying electrons and photons between network end-points, while meeting service parameters like bandwidth, latency and availability.
The agile control layer: Bringing intelligence to the network, this layer continuously configures the transport network to respond dynamically to customer service requests, assure service performance commitments and optimize the use of resources. It simultaneously organizes the transport network to more accurately meet new customer demands with its existing set of resources. At any point that it detects resources are low, it can proactively place orders for new network equipment – without even the guidance of the human handagiagilTo bring this network to life, we need five key elements:
- Artificial intelligence: Machine learning implements the agile control layer
- Programmable network: A new network, able to be configured with multiple degrees of flexibility at both the packet and optical layers
- Big data: Massive volumes of network monitoring and performance data to inform how the network is behaving
- Guidance: Top-level control to evaluate and guide the AI control layer
- Humans: Still necessary to execute tasks that cannot yet be AI-directed
In part two of this series, we’ll dive into the five key elements required for creating the agile, self-organizing networks we need to drive a new age of assurance. Stay tuned!
