AWS Wavelength is now generally available on Vodafone's 4G/5G network in London -- as the hyperscaler pushes further to the "edge" in a bid to grab an early slice of an emerging market for applications that require ultra-low latency, but which benefit from the power of the cloud: think game streaming, augmented reality, and machine learning closer to the source of data than standard cloud SaaS might allow.
The service is basically an extension of a subset of AWS cloud services into telco infrastructure; i.e. it lets users create an EC2 instance or another subset of AWS services directly in a telco provider’s network. The hyperscaler's AWS Wavelength zones currently support t3.medium, t3.xlarge and r5.2xlarge instances for mid-tier applications, and g4dn.2xlarge instances for applications such as game streaming and ML inference at the edge that require GPUs. All Wavelength services are connected over a VPC to the nearest region.
Speaking to The Stack's founder Ed Targett on the service's launch back in late 2019, AWS’s Chris Barclay (principle product manager, tech, EC2) explained: “[by embedding] AWS compute and storage services within the network operators’ datacenters at the edge of the 5G network, application traffic only needs to travel from the device to a cell tower to a Wavelength Zone running in a metro aggregation site. By creating a subnet in one or more Wavelength Zones, such as Verizon Chicago, and launching resources like EC2 instances in that subnet, we give customers the same development experience that they’re used to.
See also: Google Cloud launches Vertex AI, a new managed AI/ML platform in the cloud. Here’s what you need to know.
He added: “AWS is responsible for the service. AWS Wavelength Zones are managed by the same control plane as AWS Availability Zones.” (i.e. Customers will use the same AWS APIs that they’re used to, and won’t need to know anything about the telco provider’s infrastructure or APIs.)
AWS, which made the announcement of London availability on June 16, named six customers already using the AWS Wavelength: autonomous vehicle startup Aurrigo; sport data company Sportable: using the service to provide interactive experiences at live sports events; AR/mapping firm Immersal: tapping it to underpin immersive digital content for indoor navigation, city mapping, industrial maintenance; InterDigital: using it for smart manufacturing; Keyless, a biometric authentication startup, and video analytics/IoT firm Net4: the latter users AWS Wavelength to run video analytics.
AWS Wavelength in the UK: Partner Vodafone gets first dibs
Vodafone in June 2020 claimed to have become the first UK operator to showcase the next evolution of gigabit-capable 5G, running a a network powered by Ericsson's 5G Radio Dot system in Coventry. It also uses Ericsson as a sole network partner in London, where it has 50MHz of 3.4GHz spectrum, and Massive MIMO infrastructure set up on more than 95% of sites. It was an early AWS Wavelength partner, along with the US's Verizon, South Korea’s SK Telecom, and Japan’s KDDI.
Edge AI in particular has huge potential and is already used or being trialled within an array of sectors ranging from agri-tech to postal services. But running technologies like deep learning and natural language processing within real-world devices creates challenges.
As Daniel Warner, CEO, LGN, recently noted in The Stack, while solutions are beginning to emerge, there's work to be done before AI at the edge starts to become widely deployable.
As he put it: "Variable real-world environments, messy data, and bandwidth restrictions can turn deploying edge AI at scale into a technological and financial headache... data processing limits place natural restrictions on processing capability. With onboard memory and processing ability severely hampered by hardware, interference times lag and trainability can become a challenge."
Pointing to some of the issues that AWS Wavelength's team would no doubt argue it is aiming to tackle, Warner noted in a guest post: "Deploying edge AI as containerised cloud endpoints is inherently flawed. In real-world environments, the several hundred milliseconds it takes for a decision to go from device to cloud and back again is an order of magnitude too long. As well as incurring unavoidable lag times, edge AI in the cloud inevitably results in spiralling cloud computing costs — particularly when it comes to monitoring and reporting data." By pushing the hardware out to telco environments, AWS Wavelength aims to tackle that latency issue. The costs? It will be interesting to track performance and we'd be keen to hear from early adopters about their experiences.
