Observations from AWS re:Invent 2022

We just wrapped up AWS re:Invent last week, and what a crazy week it was. Here are some of my observations.

Attendance

• Interest in cloud computing is bigger than ever: attendance is back at pre-pandemic levels. I heard it was over 60,000.?
• As in 2019, most sessions were overflowing in each of the 3 venues and there were waiting lines almost everywhere.? As was the large busy Expo floor.

I was interested in mostly the technology updates from AWS. Here is what I took away from the major keynote sessions (VP Infra Peter Desantis, CTO Werner Vogels and CEO Adam Selipsky) and AWS as usual goes wide and deep. Vogel’s talk was broad ranging, futuristic and covered very different areas . . .? from building large complex systems from simpler computational elements, using large scale simulations, asynchronous and event-driven computing, supporting supply chain challenges to use of 3-D modeling.

Session Takeaways

• Hardware -? new hardware C7 and Graviton 3E powered EC2 is going to provide most cost effective computing esp if you are into HPC and running neural network (deep learning/ML) models. From my perspective, I don’t think your standard workloads, such as what we typically run or see with our partners, will see a huge improvement. I validated that with some of the whiteboard talks with AWS specialists and learned there are no standard benchmarks that have been published that we can reference.
• Networking -? the new SRD networking feature, which allows adaptive networking to reduce latency,? will have a much bigger and broader performance impact especially network attached services including lower latency for EBS, Elastic Cache, etc. ? This I can see as being more broadly useful. See the long tail latency benefits!??

• Serverless - AWS continues to add other features for building serverless applications. There is a nice visual composer for building with Lambda that will be useful for many AWS users. I did hear more App Dev mentioning having some Lambda in their platform services. However, monitoring of Lambda based applications was not addressed so it remains to the likes of OpsCruise to build better predictive Observability for serverless and applications. Ping me if you are working with Lambda.??
• Analytics - besides improvements in the ML area (Sagemaker), OCR (Textract), Quicksight Q for forecasting on historical data metrics, added serverless support for OpenSearch (it is open!) clusters which will benefit text and log and SIEM analytics
• AWS-only - one trend that has sustained over the years are some cloud users who are all-in on one cloud vendor; in AWS case, this is true with fintech (Capital One, FINRA) but I also met users logistics/manufacturing. I see AWS continuing to maintain this lead, unless you are into e-Commerce and will go with Azure or GCP.
• Observability - given our strong interest in this area, I attended the Observability Session (Shaown Nandi). AWS’s focus here beyond basic monitoring was more on resilience so that user workloads can respond and recover from failures. AWS has added enhanced services for DR (cross-region replication for S3), Aurora Global DB, etc. where the services can span multiple regions to reduce the huge impact past outages have created.? Unfortunately, AWS did not have anything new in its monitoring, and is limited to the traditional three pillars of metrics, logs and traces. This is probably because the focus is more on infrastructure and services, and not on user’s applications that are ephemeral and more contextual information is needed as in the case of Kubernetes.?

• Open Source - while many may not realize, AWS continues growing its support for open source technologies. Besides OpenSearch, AWS now supports many of the CNCF projects including distros for OpenTelemetry, EKS (Kubernetes), and managed Prometheus and Grafana which also includes getting logs from Loki

A final word on the Expo where there was a constant flow of traffic despite all the concurrent sessions. Pretty much every vendor that is touching or working with cloud technology was there. And as I saw in Kubecon in Detroit, and this is my confirmation bias speaking, there was a lot of emphasis on observability on the expo floor judging by the size of some of their booths.