AWS re:Invent has just completed. It was a huge event with 50,000+ attendees across 7 of the biggest hotels in Vegas, and many new service announcements.
Serverless continue to get lots of attention with new lambda enhancements and better container support. The “hybrid” model of using your own datacenter in conjunction with a cloud provider has been fully embraced by AWS with the new “Outposts” capabilities. Machine learning also got much love with several new services using and supporting it. Read on for my top picks from this week’s announcements…
1. AWS App Mesh
The most interesting announcement for me was the introduction of AWS App Mesh. It is a service mesh, allowing you to monitor and control the communication between all your microservices, utilizing Envoy Proxy and moving in to the space previously dominated by Istio.
Microservices, and distributed architectures in general, bring increased communication chattiness making it difficult to monitor, control and troubleshoot traffic. Service Meshes have increasingly been touted as a potential solution but haven’t yet gained widespread adoption. Perhaps AWS’s move into the space will bring the concept mainstream. Read more on this at
2. AWS Outposts
Outposts allow you to run AWS infrastructure in your own colo/on-premise datacenter and are a big step by Amazon to fully embrace the hybrid model that many large enterprises use. You can think of Outposts as a rack in your datacenter that contains AWS hardware.
My one concern here is what happens when the hardware inevitably fails? Presumably you have to unplug, get Amazon to ship a replacement, and reinstall. I think that is a far cry from AWS data centers which would simply auto-failover to other hardware and you likely would never know. Still, this is will be an attractive option for companies with mission critical applications running in a colo which, for one reason or another, are difficult to migrate wholesale to the cloud.
See more at https://aws.amazon.com/outposts/.
3. AWS Firecracker
Firecracker generated a huge amount of buzz at re:Invent. It’s a “virtualization technology” that allows you to create and manage microVMs, customized for serverless. That is, VMs for things like containers and functions-based services (e.g. Lambdas).
It avoids any bells and whistles to keep resource footprint, startup times, and attack surface, to a minimum. For example, the github readme reports VMs with initialization times of 125ms and memory overhead of less than 5MiB.
Honestly, I didn’t fully share the excitement since it seems like it will be used by engineers in-house at places like AWS and Docker, or those working on the Kubernetes codebase. I can’t see mere mortals like me actually using it directly. Still, it does seem like a big step forward for serverless computing and mere mortals like me will undoubtedly take advantage of it so, sure, yay FireCracker!
4. Blockchain support
While bitcoin may be a speculative bubble, the blockchain technology underlying it has great potential and AWS announced two new related services.
AWS Quantum Ledger database (QLDB) is an immutable ledger – AWS’s alternative to blockchain frameworks such as Ethereum. More specifically, it is a “ledger database”, which uses a journal that keeps an immutable, sequenced history of changes to a transaction log. It seems like this technology could have many uses cases in the financial world, such as clearing and settlement, international payments, and compliance & audit. See more at https://aws.amazon.com/qldb/
While AWS are touting QLDB as an alternative to blockchain frameworks, they did also announce a blockchain specific service, Managed Blockchain, a service which enables you to set up a blockchain network. It uses the open source frameworks Hyperledger Fabric and Ethereum.
Note that the two services can be used together: you can replicate a copy of your managed blockchain into QLDB.
5. AWS DeepRacer
The re:Invent announcemnts weren’t all dull technical announcements. DeepRacer is a “fully autonomous 1/18th scale race car driven by reinforcement learning”. DeepRacer gives you a way to get hands-on with reinforcement learning (RL), an advanced machine learning (ML) technique.
Seems like a good way to dip your toes in the autonomous driving space we will all soon and inevitably be immersed in, and you can drive with fellow-geeks at AWS Summit Events next year.
Cost: $399, available in March 2019, but you can also pre-order now for the discounted price on $249.
And I’ll leave you with some other honorable mentions from the plethora of announcements:
- AWS managed Kafka
- Amazon Timestream – a time series database
- New AWS Toolkits – improved IDE support and integration
- Amazon Forecast – time-series forecasting service that uses machine learning
- Amazon Personalize – machine learning service for individualized customer recommendations
- Amazon Dynamo DB transactions – Dynamo gets some relational goodness with support for all-or-nothing operations across tables
- Lambdas multi language support – Java, Go, PowerShell, Node.js, C#, Python were all already support in Lambda, but Ruby support was also announced and any other languages can now be supported by via Lambda Runtime API. See more at http://www.shaunabram.com/aws-reinvent-2018-keynote-announcements/#Lambda