I want to objectively take a brief look at what we’re offered by AWS for those who actually want to take advantage of 5G in their products and services, either right now or in the near future. 5G is indeed a revolutionary technology which will open brand new paths for putting all the information out there to use, it’s likely to bring the adoption of Artificial Intelligence to the very fabric of society, but I doubt if most executives out there have a clear plan for utilizing this technology with the existing tools out there. Since working with 5G will require using Cloud-based services for most companies, I think it will be a good idea to start with AWS, which is the largest of the hyperscalers.
AWS’s point of view
In his speech at AWS re:Invent 2000 on 5G, Mickey Iqbal, Solutions Architect at AWS, remarks that the most important problems CSPs need to solve for bringing applied 5G to the field are transport needs, complexity of orchestration, virtualization and, of course, monetization. According to him AWS’s solutions for those are providing microservices patterns, stateless architectures, DevOps based tooling, API based programmable architecture patterns, and containers. A bit too non-specific, if you ask me.
He says that the foremost concerns for the rest of the industry are security, connectivity, and modernization, whereas the most prominent opportunities are enhanced mobile broadband bucket, ultra-reliable low-latency, and massive machine-type communication. Below are the examples given by Iqbal as AWS’s proposed solutions for those:
- CloudFront: The fast content delivery network service of AWS.
- FreeRTOS: Including libraries for connectivity, security, and over-the-air updates for connected devices.
- IoT Greengrass: Extends edge devices so that they can act locally
on the data (such as generating predictions using ML) while using the cloud for management.
- Kinesis Video Streams: Transmits video from edge to AWS for further processing and analytics.
- RoboMaker: For creating robotic applications at scale.
- SageMaker Neo: Converts the framework specific functions for TensorFlow, PyTorch, etc. to a single compiled executable that can be run anywhere.
- Alexa Voice Service: Allows adding intelligent voice control to any connected product that has a microphone and a speaker.
- Outposts: Managed service that offers AWS infrastructure, services, APIs and tools on-premises — useful for lower latency.
- Wavelength: An AWS infrastructure which is optimized to run on the mobile edge computing network of CSPs. Wavelength zones are located within the CSP’s data center, at the edge of the 5G network. This makes it possible for application traffic to reach the servers in the zone without leaving the telecommunication network. Iqbal claims a route trip time latency of <20ms, compared to >70ms when the data goes all the way to an AWS region.
All in all, nothing especially exciting or ground-breaking here (or if there is, it’s not clearly demonstrated). Personally, I would like to see more convincing arguments to make the case that AWS has a better vision for a 5G-dominant world than its rivals.
A case study
This post explores a sample AWS architecture for network performance analytics for a mobile service provider who needs to adapt to the increased capacity that the transition to 5G will bring, especially with IoT devices. AWS services that can help with scaling are stated as follows:
- Kinesis Data Firehose: For transfer of XML measurement files that the network elements provide.
- Lambda: For transposing the XML files, converting them into CSV or JSON, and formatting the files.
- Glue & Athena: For creating an ETL job that converts from CSV to Parquet, writing the Parquet files to the object storage, and analyzing the data using SQL.
- QuickSight: For business analytics on the above data through visualizations and ad hoc analysis.
The post explains the pain points for the CSP quite well, but the proposed solutions were perhaps not that detailed and satisfying.
This is a detailed study on setting up and managing 5G mobile
network functions on AWS. The case they make for choosing AWS for this purpose includes the breadth and depth of AWS services and an API-driven approach, advantage of hyperscalability due to blooming network slicing, and its partner ecosystem across NEPs and CSPs. The reference architecture includes, along with many other services mentioned above:
- ENA and EC2 for deployment
- ELB, ALB, NLB for scaling
- Global Accelerator for network acceleration
- Direct Connect for reduced latency, cost, and increased bandwidth
- EKS for container-based microservices
- App Mesh for service meshes
- Cloud Map for resource discovery
- SNS and SQS for messaging
- CodeBuild, CodeDeploy, CodePipeline for orchestration
- CloudWatch for monitoring and performance
Again, AWS seems to identify and explain the key points for building and optimizing a 5G network quite well, but when it comes to how AWS can help with that, I find myself feeling that they are non-discriminately throwing their services at me, without much explicit guidance on how to implement them within this architecture.
We are still at the very beginning of the journey, trying to make sense of the vast possibilities that 5G will bring. I haven’t been able to see a definitive roadmap or vision by AWS but maybe it’s understandable that they don’t want to hurry too much just yet. Anyway, I certainly intend to continue exploring both the possibilities and facts of this revolutionary technology, especially in terms of AI.See you soon!