AWS Wavelength - The Impact of 5G on Cloud Services
5G technology brings significant improvement for cloud services performance by enabling services to run at the edge of 5G network. This post explains how AWS new type of infrastructure, AWS Wavelength, allows to deploy cloud services at the edge of 5G network to provide near real time machine learning inferences on the IoT data from IoT devices installed in 5G network.
With increasing number of IoT devices and real time inferencing, businesses are in need of ultra low latency applications for mobile devices. 5G network which is 10 time faster than 4G, boosts mobile application performance at the edge but still application needs to access could service and data in the cloud by passing several network hops from mobile devices to cell tower, to aggregation site, to internet and to the cloud as shown below.
This type of network increases latency and fails to provide near real time experience in latency critical IoT workloads. By deploying cloud service at the edge of 5G network as shown below, we can cut down the additional hops outside the 5G network and minimise the network latency.
AWS provides a new type of infrastructure called AWS Wavelength which offers AWS compute and storage services at the edge of telecommunications providers’ 5G networks to run workloads that require ultra-low latency over mobile networks. Location where AWS Wavelength deployed is called AWS Wavelength Zone .
AWS Wavelength zones are AWS infrastructure deployments that embed AWS compute and storage services within the telecommunications providers’ data centers at the edge of the 5G networks. AWS Wavelength zones are also connected to a Region and provide seamless connectivity to access the breadth of AWS services in that Region. AWS Wavelength zone is already generally available in some regions in partnership with telecommunications providers and AWS is actively working with many telecommunications providers to bring AWS Wavelength zone to other regions.
The following depicts how AWS resources are created in AWS Wavelength zone.
VPC - First VPC is created in the region and then subnets are created in Wavelength Zone and associated with VPC. You can create subnets in other Availability Zone and Local Zone associate them with VPC.
Subnets - You can create one or more subnets in Wavelength Zone, Availability Zone and Local Zone. Subnets created in the VPC will inherit VPC main route table which contain local route to enable connectivity between subnets in VPC. It is recommended to configure custom route table to the subnets in AWS Wavelength Zone with the default route to the carrier gateway, which allows to send traffic to the internet and telecommunication carrier network.
Carrier Gateway - VPCs with subnets in a Wavelength Zone contain carrier gateway which provides connectivity between the Wavelength Zone and the telecommunication carrier, and devices on the telecommunication carrier network. It allows inbound traffic from a carrier network in a specific location, and allows outbound traffic to the carrier network and internet.
The carrier gateway performs NAT of the Wavelength instances’ IP addresses to the Carrier IP addresses from a pool that is assigned to the network border group. The carrier gateway maps the network interface (EC2 instance etc.) IP address to the Carrier IP address, and then sends the traffic to the telecommunication carrier.
Carrier IP address - This is the address assigned to a network interface which resides in a subnet in a Wavelength Zone. Carrier IP address are allocated from a network border group which is unique set of Availability Zones, Local Zones, or Wavelength Zones from which AWS advertises IP addresses.
AWS allows to create the following resources in AWS Wavelength Zone.
- Amazon EC2 instances
- Amazon EBS volume
- Amazon VPC subnets
- Carrier gateways
- Amazon EC2 Auto Scaling
- Amazon EKS clusters
- Amazon ECS clusters
- Amazon EC2 Systems Manager
- Amazon CloudWatch
- AWS CloudTrail
- AWS CloudFormation
You can create the above mentioned AWS resource in AWS Wavelength Zone and deploy the application / service to provide ultra-low latencies to mobile devices and end users over the 5G network. 5G connectivity along with AWS Wavelength brings new paradigm shift in providing near real time machine learning inference at the edge for smart industry automation, IoT, and autonomous vehicles.
Let’s consider a use case.
To stop the outbreak of COVID 19, Company X is installing COVID 19 IoT sensor (5G device) in public places which will collect and send air quality data to machine learning inference engine to detect the existence of COVID 19 virus in near real time.Once detected, we have to activate the alarm in sensor device. Assume we have already trained the machine learning model using past 2 years air quality data to detect existence of COVID 19 virus in the air. Company X expects the number of sensors will increase rapidly and the solutions should be highly scalable, resilient and secure and need high availability and low cost.
Let’s see the solutions for the above case. Here, sensors are 5G devices which will connect to 5G network and the inferencing should occur near real time. Therefore, we are after ultra low latency solutions which force us to deploy the machine learning model in 5G network. Besides, system should be highly scalable, secure and reliable and needs high availability and low cost. Thus, we need to deploy the solutions in a cloud infrastructure. These requirements lead us to have cloud infrastructure in 5G network. AWS partnered with Telecommunication providers and offer AWS Wavelength Zone on 5G network. In order to fulfil Company X requirement, we are happy to deploy the machine learning inference engine in AWS Wavelength Zone as shown below.
- Inference engine is deployed in AWS ECS environment AWS Wavelength Zone.
- Sensor will frequently collect air quality data and make call to inference engine.
- Air quality data and inference engine response will be stored in S3 bucket for training and tuning the machine learning model.
- Based on inference engine response, sensor will activate the alarm.
In this post, we covered why we need cloud infrastructure in 5G network, AWS Wavelength architecture and implementation of near real time inferencing solutions using AWS Wavelength.
