Artificial Intelligence (AI) has received a lot of hype for years now, and more is coming in the telecom domain because AI and streaming analytics have become central to operators’ success with 5G due to the increased complexity from a growing number of connected devices, a more diverse multivendor ecosystem, significantly higher data volumes, and the distribution of data from their 5G network edge to the core to cloud. AI-based streaming analytics are essential for 5G automation and your ability to deliver new 5G services with quality customer experience.
But how specifically will 5G and streaming analytics change things for you in 2022? Here are my top 3 predictions:
#1: 5G will nearly eliminate the use of batch data ingestion techniques for capturing telecom data
5G is deployed on a vast number of smaller cellular sites, and those deployments all have CPUs. It’s wasteful not to take advantage of that distributed CPU power, and a great way is to perform data cleansing, aggregation, sessionization, and other streaming analytics at the edge before RAN data is ever transmitted to the 5G core. This has the added benefit of reducing the network bandwidth and storage requirements for 5G RAN data because sessionized RAN data is 10x to 100x more compact than raw data.
#2: 5G will push the data processing architecture in telecom to the edge
5G generates substantially more data than previous telecom network architectures. Simply moving all that data to a central location and processing it in a traditional manner will fail or become cost-prohibitive due to volume.
In addition, new 5G applications require much lower latency to analytics than ever before, which in turn requires the implementation of streaming analytics. Consider, for example, smart urban traffic light timing. Anyone who uses Waze, Google Maps, or Apple Maps can easily understand the large volume of data collected. It is not a stretch to imagine a world in which all that data which tracks vehicle positions and speeds in real time is used to change the timing of traffic lights to move cars more efficiently through a metropolitan area. But, doing so requires analytics in sub-second time; otherwise, the traffic light timing changes won’t reflect the true state of the road network and therefore won’t capture the opportunities.
Drone management flight planning is similar. What happens when a network failure occurs? Flight replanning to keep a drone in network coverage must happen in real time; otherwise, a drone can lose contact and be forced to return home or hover indefinitely. And, either of those stopgap measures could lead to a drone that runs out of battery power.
Many other applications have similar characteristics. In the world of drone management, signal modeling, noise management, flight plan fairness, and anti-drone operations detection all require real-time response. In smart energy management, anomaly detection, equipment monitoring, predictive maintenance, and energy demand forecasting are also all real-time questions.
Adding the latency associated with transmission to the 5G core and traditional big data processing architectures will fail to meet the latency requirements of new 5G applications. This will force telecom deployments to embrace streaming and edge data processing architectures that also provide the benefit of lower latency to analytics.
#3: 5G will spawn a new class of analytics applications that run in the RIC
Prior to Open RAN, entrenched telecom suppliers had no incentive to develop architectures in which the radio were – say – 50% more efficient. Why would they? More efficient radios mean fewer radio sales and therefore less revenue. But with Open RAN, the open architecture allows third parties from the telecom ecosystem - not just radio vendors - to innovate with new applications that run inside of a 5G RAN Intelligent Controller (RIC). Examples such as autonomous vehicle (V2V and V2I), augmented reality, virtual reality, multiplayer gaming, network quality of experience optimization, network quality of service-based resource optimization, and radio access network slice service-level agreement assurance require responses to data in real time, where real time is defined as less than a second and in some cases less than a millisecond. Stream processing and streaming analytics technologies can now operate below that threshold making such applications possible.
To take advantage of these 3 big areas of change, here are some key actions you can take this year
- Measure the performance of your data processing pipelines in flows per second per core as well as overall application latency to understand how much hardware is required to implement applications at 5G scale.
- Update your application architecture to use both stream processing and streaming analytics techniques and technologies, especially where the required latency is very low and the total required core count would otherwise be larger than feasible.
- Aim to reduce your total cost of ownership by shifting as much processing to the edge as possible, avoiding unnecessary storage and network utilization costs.
With that in mind, I wish you a prosperous new year filled with the joy of implementing your 5G streaming analytics solutions! And if, for any reason, Waze, Google Maps, or Apple Maps gets stuck calculating your route, be sure to let your human brain take over the navigation process because you won’t be able to sit indefinitely on the freeway until the service finally responds.
