Stories

AWS re:Invent 2022 - Adam Selipsky Recap

Written by Jeremy Nees | 30/11/2022 1:20:00 AM

With a tinge of jealousy, I’m sitting on the couch this morning watching the keynote for Re:Invent 2022 as many Kiwis flock back to Vegas for the first time in a long time. Some are even seemingly completing the final leg by private jet (WTF. Since when was that a thing!)

In typical keynote fashion, Adam Selipsky takes the stage to a band - this year bashing out the main riff to Sweet Child of Mine. And the CEO's Re:Invent 11 keynote is underway. 

First, the stats come in hot and heavy.

There are more than 1000 unicorn startups worldwide and more than 80% run on AWS. Think DataRobot, Whizz.

AWS is the world's largest purchaser of renewable energy. And is targeting 100% renewable energy by 2025. And water positive by 2030 - I didn’t really know that was a thing…. I guess it’s a thing.

We’re also reminded that you can turn cloud resources up and down, which comes in pretty useful if you’re an Airbnb.

Reading between the lines, AWS is big and impactful and they’d like to make sure we don’t forget it. Duly noted Amazon.

And that's the introduction out of the way. On to the announcements.

So to get into it we start with some Star Trek stuff. I feel an analogy coming. With the naked eye, we can see 2.5M light years to the edge of our galaxy. I’m waiting for “the final frontier”. We look at the Pillars of Creation picture and different views or filters. Lots of stars. Lots of ways to look at them. And there it is! We’re talking about data.

Announcement: AWS Open Search Serverless.

The first announcement is that a serverless option is now in preview for Open Search.

Maybe the most important part of the announcement is that all AWS Analytics Services will now be available Serverless.

We whip through a customer case study with Engie, an 180 year old energy company that tells us global energy demand will grow to 270000 TWh (50%) by 2050. Yikes, if I could put that in perspective.

Back to the announcements!

Announcement: Amazon Aurora zero ETL Integration with Redshift.

Tables in Redshift in seconds with no data pipeline. All serverless.

Now, this is potentially quite big. Depending on where AWS takes it. There are a number of third-party products that create virtual data lakes of sorts, allowing you to present data through a single catalogue with minimal requirements to move it from its source. See Zetaris as an example. One downside with modern data lake architectures is that we are often replicating a large amount of data, just to organise it centrally and to simplify what we can then do with it. AWS’s announcement only applies to a few data sources within the AWS ecosystem, however, it wouldn’t surprise me to see this expand beyond the AWS ecosystem. Of course, there are pros and cons to each approach, largely dependent on how much transformation, enrichment or data manipulation is actually required and performance characteristics.

Announcement: Redshift integration with Apache Spark.

Analyse data in Redshift using Spark queries. I don’t have anything really to add…

We now go into a bit more commentary about the balance between Control and Access for data governance. Too much Control and you wind up with a lack of productivity and shadow IT. Too much Access and your data will end up where you don’t want it. Establishing the right governance creates trust and confidence for innovation. And on that note:

Announcement: Amazon Datazone allows you to set data free safely.

Enable data scientists, product managers etc to collaborate around data. Datazone is driven through a web UI to create a data catalogue and connect it to data sources and then enable access to the catalogue by consumers. Simplifies the whole creator and consumer side of the data equation. Good job I say!

We continue to talk about data, shifting focus to business intelligence. If you recall at the last Re:Invent, we were introduced to Quicksight Q which added natural language processing to the AWS Quicksight BI tool. Did I mention this is powered by ML? Of course, it is…

Announcement: Quicksight Q - Machine-Learning-based forecasting.

So we know we can ask Q natural language questions about the past, but now we can ask questions about the future! An example would be sales forecasts. Q also now has “Why” questions, as in “Why did you just tell me XYZ”. Q will then tell you how it came up with the set of possibilities and the confidence in those answers.  

Guess we no longer need this sports fan:

We’re on to another analogy, this time about exploring the ocean floor. Only 30% of the ocean floor has been explored. Oh! We need the confidence to explore - like IT engineers - and AWS gives you that. So we’re talking about security now.

Announcement: Container runtime protection for GuardDuty.

A nice addition to the GuardDuty family. There are lots of point products around for container security. But yeah, point products. While they can be awesome, a lot of them become hard to manage. This sounds useful.

And then what I would call the big one for security:

Announcement: Amazon Security Lake allows security teams to collect, combine and analyse at a petabyte scale.

Supports the OCFS from third parties and internal AWS sources. Automatically handles retention and a few other things.

So we pretty much just said AWS SIEM without saying SIEM (and I’m sure someone will contact me to tell me why this isn't a SIEM but a Security lake. If this is you, just accept I’m a simpleton). Of course, the other hyperscalers have their plays here, with Sentinel and Chronicle, with this being something that has been strangely lacking from the AWS ecosystem. Yes, partners provide lots of solutions that run on AWS, but I can see some benefit in a native solution here, especially if you’re all in on AWS and you have a lot of log sources within the platform. 

Of course, you need a way to analyse this data and both AWS tools such as Athena, and OpenSearch are up to the task as are some ecosystem tools such as Splunk, IBM Qradar, Sumo Logic and VMware Cloudhealth.

So moving on. 

We are getting into the chipset announcements. And I refuse to do it and talk about the details. Instead, I will focus on the fact that it seems a near requirement now for US tech execs to wear blue jeans with white-soled sneakers. 

My summary. Things are faster and cheaper and more efficient (you knew this right). A few things in here around HPC - clearly something of a focus and where cloud can be leveraged with the right architecture - remembering that HPC has never been about what one compute unit can achieve but more about how these are interconnected and access storage to allow for parallelisation.

Probably the overarching theme is that while virtualization traditionally created a generic virtual hardware abstraction from a raft of often mismatched generic hardware, at the scale of the cloud, it is increasingly beneficial to have custom silicon that is optimised for particular workload patterns and that the hypervisor is able to pass the benefits to the consuming workload. Enough on that. 

The next scene setter is all about Imagination, with a nod (unsurprisingly) to the creativity behind the Lord of the Rings. And the message is that creativity thrives in a collaborative environment. We are now into the final run and some more vertical-specific products and solutions. 

Announcement: AWS SimSpace Weaver.

Large-scale spatial simulations without scaling infrastructure. Design in a 3D engine then upload code and execute in a few clicks.

Announcement: AWS Connect.

New capabilities with three new features for Connect 

  • ML forecasting for resourcing, scheduling etc
  • Agent performance management analysis
  • A new UI that guides agents through customer interactions so they can resolve issues faster 

Announcement: AWS Supply chain.

Connect supply chain data, AWS supply chain automatically sets up a data lake with pre-formed ML models to create a unified data model.

AWS supply chain then contextualises this data in a supply chain map.

You can create watchlists which will alert on risks such as your apples being overstocked or understocked in a particular location. It then provides recommended actions such as redistributing stock from one location to another. This is quite cool. I like it. 

Announcements: AWS clean rooms.

Securely collaborate with third parties on data without exposing all your data. Reads data where it lies and applies privacy-enhancing controls. 

Announcement: Amazon Omics

Purpose-built services to store and analyse genome and other omics data. Also integrates with Amazon Health Lake allowing analysis across data sets. 

And finally, we finish with Amazon Just Walk Out - and pay with your palm of course with Amazon One. Enabling retailers to provide an experience where you can walk into a store, grab your stuff, swipe your palm and then head out. I think this will take some adjusting to for consumers, but ultimately we are hardwired as humans to expend minimal energy and therefore it will happen. Read a blog on it here: https://aws.amazon.com/blogs/industries/tag/just-walk-out-jwo-technology/

Well, that’s it from today's keynote. The key themes I’d summarise as data-oriented and increasingly vertical-focused solutions. I expected to see a bit more about Low Code/No Code given how much this featured last year but we may see that in the coming days.