Taking the Blinders off 5G
Today’s service-blind networks will need a big change to manage the demands of a 5G world where more users, more data and more services are just the beginning. 5G is upping the ante for carriers and service providers as each user or device will potentially require a different type of service – something our current way of provisioning and managing service level agreements (SLAs) does not take into consideration.
What do we mean when we say the service type criteria changes from user to user? For example, this can mean guaranteed bandwidth 24/7 or only at a limited scheduled timeslot, maybe super-low latency, or even 5 9s availability. This means the service offered also must be able to adapt from both time to time and situation to situation – dynamically, and fixing issues in real time.
We’ve been through the progression from 2G to 4G, but this “G” is different. For instance, 5G will be the first network designed to support non-telecommunications services, to bridge wireless and wireline technologies, and to provide new ways to support both human and massive machine type communications. In a nutshell, 5G is requiring us to completely rethink network provisioning and traffic management.
Providers can no longer to rely on (static) over-provisioning that worked in 3G and 4G to ensure that the new service type guarantees are met. To deliver as promised, means these types of services will need to be dynamically provisioned, optimized, and assured. It will also mean dynamically adjusting the network itself to traffic surges that are often caused by external events (like gaming, sports, world news, weather, etc.) or other failures.
Moreover, taking a “Scout’s Honor” approach to delivering on an SLA will no longer be sufficient in 5G. This will necessitate a proactive approach to service assurance. Imagine important vehicle-to-vehicle (V2V) communications degraded just as thousands of autonomous vehicles are caterpillaring down the freeway, practically bumper-to-pumper, during rush hour? Or how about a slight delay in the feed for a remote surgery? Penalty avoidance based network assurance of the past will no longer be acceptable, because customers require assured services and are not interested in collecting penalties.
The varied nature of the user or device and exterior factors puts the greater pressure on monitoring performance and preventing service lapses – something current tools are not designed to do at this heightened level. Nothing currently available allows vendors to monitor end-to-end performance, let alone correct issues and reroute traffic in real-time.
The rise of artificially intelligent, multi-gear networks
A recent concept we explored in the February edition Connect-World that will be distributed at Mobile World congress, we explore the topic of building artificially intelligent, multi-gear, cognitive networks to deliver service assurance for 5G and beyond. The notion of truly cognitive networks, may not be new, however, the reality of putting this in action will face many challenges.
As networks need to be dynamic, multi-faceted , slicing enabled and assured, it is clear that there is no one technology that will enable carriers to cost effectively deliver all of the promised services and network slices in real time, and with competitive prices. Just like cars have gears to support dynamic driving situations, network devices to will need to be “multi geared”, to concurrently support multiple data and control planes technologies (IP, MPLS, FLEX Eth, OTN, WDM etc.), and be able to route different services and demands on the right slices on the right time.
Multi-gear networking is essentially the ability to combine packet and optical technologies in various ways to create network slices to enable the efficient transport of different service types. And multi-gear networking’s built-in flexibility allows resources to be dynamically shifted from one network slice to another, based on the creation of new services and changing traffic patterns. This in turn will dramatically increase the factors and parameters needed to be taken into account in service creation. And thus support of 5G services will require controlling and managing a multitude of constraints related to the network, and humans’ ability to process these factors and then act in real time, or at any time, will be nearly impossible. However, achieving cognitive networking right now is impractical for the same reasons mentioned: it isn’t viable to create the calculations needed to ensure the network functions as needed, to reach every SLA specification, with the network management tools we have today.
This brings us to artificial intelligence for multi-gear networks. As mentioned, humans are not able to process and take actions on the multitude parameters and constraints associated with 5G services on multi geared networks. However computers, specifically AI based applications, can do that very efficiently and even learn and correct their algorithms in real time. By harnessing emerging technology like artificial intelligence and machine learning, we can achieve the necessary self-organizing capabilities required in 5G networks. This includes real-time awareness of service needs and network capabilities while autonomously and continuously improving network slicing provisioning, monitoring and assurance decisions to ensure SLAs are met.
With 5G, these AI-based tasks will be commonplace, particularly in the realm of IoT, smart cities, and intelligent transportation systems.
Dynamic, optimized, and assured networks?
When it comes to dynamic, optimized and assured networks, most carriers address only two of these three, and in many cases, one or even none. However, for 5G, all three will be needed to achieve truly dynamic, optimized and assured network. But doing so, service providers will need to leapfrog current capabilities. This means, already expensive 5G networks will require even more investment.
An artificially intelligent, multi-gear network is a big departure from today’s traditional networking. Fortunately, they can be introduced in phases. The initial phase, for example, can cover a limited geographic area or a specific portion of the network, like the backhaul between the radio and the network core. Additionally, the AI engine can mandate human oversight or limit autonomous decision making. In this manner, we can gain experience and confidence in an AI-driven multi-gear network on a manageable scale and then expand deployment in multiple dimensions.
The main point to remember, carriers cannot just build thicker pipes to meet SLAs in a 5G world, especially if they want to see a return on investment. Carriers need to sell bandwidth resources several times over, dynamically and for different types of services, (e.g., Network-as-a-Service (NaaS)). This requires a fundamental change in the way service providers think about services, especially if the goal is for additional, flexible revenue streams, and, more importantly – service assurance.
This next phase is exciting and challenging. Our approach to helping service providers make a shift with the future in mind, including developing artificially intelligent multi-gear networks, will be one of the many topics we’re be discussing at Mobile World Congress in Barcelona, Spain. If you’re around the show, come see us in Hall 6, Stand C610.