#729: AWS News: Aurora Storage Upgrades, DynamoDB Multi-Region Strong Consistency, and More

0:00This is episode 729 of the AWS podcast, released on July 14th, 2025.

0:09Hello, everyone, and welcome back to the Adabas Podcast. I'm Lesh here with you. Great to have you back. All three availability zones in effect today. Firstly, welcome to the podcast, Jillian Ford. Welcome. It's always a good day when it's a podcast day, Simon. It is a fun day. And joining us for this fun day is another of the availability zones, Shruti Kopaka. G'day, Shruti. How are you doing? Hi, I'm great. You know, we celebrated a long weekend in the US this past weekend. So I am very dejuvenated and ready for this. Both of you are very well rested. It's glowing, yes. Nothing like a long weekend.

0:46Well, lots of updates. In fact, there's 68 updates today to go through, but a couple that really caught our eye. firstly, Amazon Aurora PostgreSQL database clusters now support up to 256 TB bytes of storage volume. Now, that's a lot. This is doubling the previous limit of 128. So as I always remind people, the only reason these limits go up is because people need the limit to go up. So it's always interesting to me that folks have big databases. But I remember when this was like an unfeasible database size, but now this is normal. Yep. And I mean, what's really great is that this, it doubles the capacity, but it still kind of maintains its cost efficiency because you're still doing pay-as-you-go model.

1:37Customers will still only pay for what they use, even if they upgrade to this new database version. And yeah, I mean, as you said, It's really good for customers who have these really large storage requirements, but also like historical data retention requirements. And they had to do all this themselves. Now it's just scales to their needs. So that's cool. And the funny thing, I just was looking through some of the things that we can still have up to 15 read replicas with this database. So that's a lot. It's like we're now in petabytes. Wow. Yeah, that is. That's a lot. That's very cool. Now, speaking of, I'm all about databases today.

2:21Speaking about another database improvement that's taken place is back at reInvent, we talked a lot about the new consistency modes for DynamoDB, in this case, multi-region strong consistency. That's now generally available, which I think is really interesting for a lot of customers, Julian. I think so too. This one I'm super excited about. I think as people have always thought historically about no no sql databases um and then we've always thought about like consistency being something that maybe you'd think about as a trade-off and now with dynamo db and it's almost like you you don't have to necessarily make that trade-off now um because of this feature and multi-region multi-region strong consistency i mean that is absolutely amazing it's hard but it's done.

3:13Now you've got to be, you have to be in three regions. You can either be in three regions or you can have two replicas and a witness, but three is the number and the number shall be three. And, uh, I'm sure that many people will get that particular reference. Um, but my two co-hosts are clearly not Monty Python fans. So they didn't get that reference. They're not nitty enough. That's okay. They're looking at me confused, everyone. I haven't talked about mind pretty well in a while. I had to think about that. Yeah. That was a very, very Monty Partham, the Holy Grail reference there. And I know my audience and I'm pretty sure my audience will appreciate that.

3:48I think so. So I'm no database expert, but so explain to me what this like strong consistency is. Is it basically like, for example, if there's a financial service company who say needs to be able to show the real-time account balance, regardless of where it's accessed from, like whether it's accessed from Tokyo or whether it's accessed from London. And there was a transaction that happened in Tokyo which updated the balance. So in real time, all the copies across all the regions will be sort of updated so that, you know, customers who like financial services customers originally would have had to like build all this complex logic themselves, but now it just does it.

4:31I think that's a very, very good, very good assessment of what it is. Basically, and I'll read from the documentation to be specific, item changes in a multi-region strong consistency global table replica are synchronously replicated, so that's synchronously, not asynchronously, are synchronously replicated to at least one other region before the write operation returns a successful response. Strongly consistent read operations on any replica always return the latest version of an item. Conditional rights always evaluate the condition expression against the latest version of an item. So yeah, you have a consistent view across the world.

5:14Yeah, that's pretty magical. Yeah, it is. Well, let's talk about some other updates, and they've been a lot. So firstly, with analytics. Adibus Glue enables enhanced Apache Spark capabilities for AWS Lake Formation Tables with full table access. So this means that you can use data manipulation language like create, alter, delete, update, and merge. So this means you can have a lot more control over what's going on. Amazon Redshift Serverless now supports four RPU, that's a Redshift processing unit, minimum capacity options. So it's smaller. It's a smaller data warehouse, which is useful for a lot cases, a little bit more granular.

5:53Amazon QuickSight has launched trusted identity propagation or TIP for Athena direct query. With this capability, customers can apply fine-grained access controls using lake formation rules to govern user access to data in QuickSight. So TIP allows authors to securely control rows and columns of data returned by queries, allowing the same dashboard to be used across customers or departments. This is important because often you should only see the data you should be able to see, but the queries may be the same for everyone else. AWS Cleanrooms now support two enhancements to its machine learning capabilities to let you train your models more efficiently and to scale.

6:35So incremental training enables you to build upon existing model artifacts to create new models. And distributed training allows you to train models across multiple compute instances simultaneously. That's pretty cool. adibus announces that amazon sage maker has contributed a custom transport amazon data zone transport to the open lineage community and enhanced automated lineage capabilities now lineage is important because this explains where data came from this has become more and more important in this amazing new world of generative ai so these lineage enhancements include improvements to automation from sources like adibus glue amazon redshift and automated lineage capture from tools, which means you can work more efficiently and understand where the data came from.

7:23Quick updates for application integration. Amazon EventBridge now supports AWS CodeBuild batch builds as a target, so you can trigger off EventBridge activities. AWS Security Incident Response has also added integration with Amazon EventBridge, so this lets you react, monitor, and orchestrate events associated with cases and memberships within your response. We're excited to announce Qindex now supports seamless application level authentication for its search relevant context API which makes the end user experience when using this index to retrieve content easier. And AWS service reference information now supports annotations for service actions.

8:05Now action properties provide context to indicate what an action is capable of such as write or list capabilities when you use it in a policy. Now, service reference information streamlines the automation of policy management workflows because you can retrieve available actions across AWS services from machine-readable files. This is a big deal for the automation-minded amongst us. Let's talk artificial intelligence. Amazon Nova Canvas has added a pretty cool feature. this is a virtual try-on and style options for image generation. So you can now see what clothing would look like on a shopper, or you could see furniture intelligently placed in a living space.

8:48You just upload the two images, the person or the space and the thing, and it puts them together, which is kind of interesting. So you can try on all kinds of clothes. I can see people getting up to all sorts of mischief, trying on different clothes that they would never wear. Citations, API and PDF support for Clawed models is now available in Amazon Bedrock. Now the Citation's API allows Clawed to ground its answers in source documents, providing detailed references to the exact sentences and passages used to generate responses. This is a really interesting and important capability because often when you're analyzing a big document and you're extracting, in fact I have an application I've written to do this for another purpose, and it does exactly this capability.

9:31So I'm a bit excited about this. It's what I love about this show is I get to read through all these updates and go, oh, I'm going to use that. Amazon Bedrock Flows announces preview of persistent long-running execution and inline code support. So Bedrock Flows customers currently get three limitations when authoring, executing, and monitoring their workflows. A two minute idle timeout restriction per step, the need for custom API monitoring solutions, and the requirement to create Lambda functions for basic data processing tasks. Sounds like some undifferentiated heavy lifting, so starting now we're addressing them with some new preview features that extend workflow step execution to 15 minutes and let things run much more easily for you.

10:25AWS Health Imaging has launched support for DICOM Web StoRS data imports. And in a classic case of if you work in the health industry, if you know, you know. And so now there is a synchronous data import action that's ideal for latency sensitive workloads, like storing new medical image studies, annotations and reports. Amazon Recognition Face Liveliness has launched accuracy improvements and a new challenge setting for improved re-reporting. Amazon Recognition Face Liveliness has launched accuracy improvements and a new challenge setting for improved UX. So this allows you to figure out if the face you're seeing is a real face.

11:06And there's a whole bunch of nuance in doing that that I'm not going to go into now, but they've improved it. Amazon Textract has announced accuracy and feature updates to detect document text and analyze document text APIs. So it is now even better in terms of what it does. And I know a lot of customers who use this all the time. So distinguishing between the letter O and the number zero on a fax, a low-resolution fax, not easy to do. So they've done some work on that. Amazon SageMaker Catalog has added AI recommendations for descriptions of custom assets. So this reduces the manual documentation effort, improves your metadata consistency, and accelerates asset discoverability across organizations.

11:48And we're also happy to announce Amazon SageMaker HyperPod training operator. This is a purpose-built Kubernetes extension for resilient foundation model training on HyperPod. Amazon SageMaker HyperPod empowers customers to accelerate AI model development across hundreds or thousands of GPUs with built-in resiliency, decreasing model training time by up to 40%. Now, this resiliency is really important. When you talk to folks who are doing this model training, they'll tell you that the biggest problem they have is stuff not working through a long-running job. And as training clusters expand, this is only getting more disruptive.

12:26Value recovery traditionally requires a complete job restart across all nodes, even when a single training process fails, which means you've got more downtime and more cost. And figuring out critical training issues like stall GPUs, low training throughput, put numerical instabilities, often needs a lot of custom monitoring code. With the HyperPod training operator, customers can further enhance training resilience for Kubernetes workloads. So instead of a full job restart when failures occur, the HyperPod training operator performs a surgical recovery, selectively restarting only the affected training resources for faster recovery from faults.

13:03It also introduces a customizable hanging job monitoring capability to help overcome problematic training scenarios like store training batches, non-numeric loss values and performance degradation through simple YAML configurations. So you can get up and running nice and quick, but this is really, really useful if you use SageMaker HyperPod. And AWS Health Imaging now supports DICOM Web bulk data. So again, more improvements in the processing of medical information. And AWS Health Omics has announced automatic input parameter interpretation Oops. And Ademis Health Omics has announced automatic input parameter interpolation for NextFlow workflows.

13:47And finally on this one, Amazon Q Business has launched the ability to customize responses. So with response customization, customers can provide instructions for identity, tone, and output style when configuring your Q Business application. So this really helps you adjust the chat persona. its communication formality, response length and detail. And it also has built-in guardrails to ensure that response settings align with your existing admin controls. Up next, we've got business applications. And usually this segment is just about Connect. But we've actually got some other updates as well. So this one's going to be exciting.

14:29Amazon Q in Connect now supports seven languages for proactive recommendations. Those are Spanish, French, Portuguese, Mandarin, Japanese, Korean, and English. Amazon Connect now provides an enhanced flow designer UI for making it a lot easier to be able to edit. So these are things like keyboard navigation, screen reader support, all of the good things that just make it a lot easier for you. Amazon Connect can now integrate agent activities from third-party applications as Connect Tasks, which can be evaluated alongside work completed in Connect, providing managers with a unified application for quality management.

15:14Amazon Connect now provides the ability to execute logic such as routing, priority changes within a flow, while continuing to play audio to customers who are waiting in a queue. And this is a new one. I don't think I've ever done an update on this service, so I'm going to actually have to do some background. And it's called AWS B2B Data Interchange. And I will admit, I didn't know what this service was. I know I shouldn't say that out loud. Well, you're not old enough to understand what it's even responsible for. Because when we're talking about X12 EDI documents, let me tell you, we're going back to the 1990s.

15:54But this stuff still exists out there. And it's cool. Well, this is the 90s theme episode. So we've got Monty Python and electronic data interchange documents. And that's what this service is all about. AWS B2B data interchange. This automates the transformation of business critical EDI transactions. So that could be for healthcare documents, financial documents. It's just going to make that a lot simpler. And there's actually a few updates here. So there's going to be new configuration capabilities. These are going to allow you to have some customized formats and technical fields to be able to generate some of these EDI documents.

16:37That's going to make it a lot more compatible when you're working with additional trading partners. So some of that could be like documents that are in ANSI X12 documents. And if you need to get those from JSON and XML data formats, that will be a lot simpler. We also had another one from the service as well. AWS B2B Data Interchange now supports splitting of inbound X12 XCI documents that contain multiple transactions into individually processed single transaction documents. And back to a couple more updates from Amazon Connect because they always steal the show on the business section. They're so busy all the time.

17:21They're just, I don't know. We need to meet some of those software developers. They're just killing it. I totally agree. So customer profiles, this now enables organizations to create customer segments from imported CSV files. So now contact center managers and campaign managers, they can upload predefined customer lists. And this is going to streamline the process of when you have to do targeted segments for personalization. There's also some launches on the API sides from Amazon Connect. This allows for you to be able to delete cases, case comments, if you need to undo contact associations, and remove service level agreements from cases.

18:10Another one from Connect, it now supports custom work labels for agent schedules, And this makes it easier for you to identify the type of work an agent is scheduled to be able to work with. So some examples of this is you could add order processing as a work activity for Monday, or you could add returns management for Tuesday. Those are just a few examples that can really help simplify the experience for managers. And then we've got one update here on Amazon QuickSight. So this now supports 2 billion, with a B, rows in the SPICE dataset. So if you're not familiar with the SPICE acronym within QuickSight, it stands for Super Fast Parallel In-Memory Calculation Engine.

19:01That's the magic that Amazon QuickSight uses to make all of your pretty dashboards load really fast because they are in memory. All right. Well, next up, we have quite a few updates under Compute. AWS announces availability of ECS-optimized Windows Server 2025 AMIs, or the Amazon Machine Images. They offer two distinct platforms, 2025 core and 2025 full. These AMIs are specifically engineered to support Windows container deployments on Amazon ECS or our Elastic Container Service. Each AMI comes ready to use with essential components and optimizations for running containerized workloads and makes it really easy to deploy these containerized workloads.

19:56AWS Fargate now supports Seekable OCI or SOCI Index Manifest v2 for greater deployment consistency. SOCI or Seekable OCI accelerates Amazon ECS task launches by enabling containers to start running before the full container image is downloaded. This index manifest v2 uses a cryptographic method to establish an explicit link between the image and its manifest, ensuring integrity and consistency during and across all the deployment stages. The next one is an especially exciting one. Amazon EC2 C8 GN instances are now generally available. So these are the new instances powered by our Graviton 4 processors, and they provide up to 30 % better compute performance than the Graviton 3-based Amazon EC2 C7GN instances.

21:03C8GN instances feature our latest generation, which is, I think, V6, so sixth generation of the AWS Nitro cards, and they offer up to 600 gigabits per second of network bandwidth, the highest network bandwidth among network-optimized EC2 instances. So customers can take advantage of CAGN instances and their enhanced networking capabilities for network-intensive workloads, such as network virtual appliances, data analytics, any kind of CPU-based AI and machine learning inference, and much more. Also, for increased scalability, C8GN instances offer instance sizes up to 40x large with up to 384 gigabytes of memory and up to 60 gigabits per second of bandwidth to Amazon Elastic Block Store.

22:06So a lot of powerful compute networking and sort of storage capabilities packed into these instances. Go and check those out. Next up is an update on being able to use Amazon EBS GP3 volumes with AWS Outpost Racks. You can now use these GP3 volumes with second generation of AWS Outpost Racks for your workloads that require local data processing and data residency. With GP3 volumes, you can scale up to 16 ,000 IOPS or input-output operations per second and 1 ,000 megabytes per second, delivering 4x the maximum throughput of the previously supported GP2 volumes. Amazon ECS includes task ID in unhealthy service events.

23:06This makes it easier to troubleshoot unhealthy tasks by adding the task ID in service action events generated due to health failures. AWS announces Ubuntu support for Finch, an open source command line tool that allows developers to build, run, and publish Linux containers. Finch simplifies container development by bundling a minimal native client with the curated selection of open source components. Now with the addition of Ubuntu support, Finch provides a consistent and streamlined container development experience across more Linux distributions, addressing a key pain point for developers who are working across multiple environments.

23:48Next up is an update to AWS Neuron, which is the software development kit for AWS Stranium and Inferentia, which are the purpose-built AI chips AWS has built. We are announcing the general availability of Neuron 2.24. Neuron 2.24 announces support for PyTorch 2.7, enhanced inference capabilities such as prefix caching for faster time to first token, as well as disaggregated inference that reduces the pre-fill decode interference and contextualism for improved performance on launch sequences. So that was it for compute and moving on to databases. Amazon Relational Database Service Custom or Amazon RDS Custom for Oracle now supports multi-AZ deployments.

24:47This will now provide high availability for business-critical workloads across, of course, so many different regions. With multi-AZ deployments, RDS Custom for Oracle synchronously replicates data between two availability zones and performs an automatic failover in case the primary database instance becomes unavailable so that customers will benefit from this higher availability. Amazon Aurora MySQL and Amazon RDS for MySQL integration with Amazon SageMaker is now available. This integration automatically extracts and loads data from MySQL tables into your Lakehouse, where it's immediately accessible through various analytics engines and machine learning tools.

25:38The data synced into Lakehouse is compatible with Apache Iceberg Open Standards, enabling you to use your preferred analytics tools and query engines such as SQL, Apache Spark, BI, and AIML tools. Amazon Aurora now supports PostgreSQL 17.5, 16.9, 15.13, 14.8, and 13.21. The update includes PostgreSQL's communities product improvements and bug fixes while introducing Aurora-specific enhancements, which are read replica optimizations to reduce downtime during cluster upgrades, new features for Babelfish, and some security improvements. Amazon Keyspaces for Apache Cassandra now supports change data capture streams.

26:36CDC streams or change data capture streams in Amazon Keyspaces automatically capture insert, update, and delete operations as change events, delivering them in order with automatic deduplication. With these CDC streams, you can build event-driven applications and implement use cases such as data analytics, text search, machine learning training and inference, and continuous data backups for archival. Amazon Neptune Graph Explorer introduces native query support for Gremlin and OpenCypher. This enhancement empowers data scientists, developers, and database administrators to seamlessly interact with their graph databases using their preferred query language, eliminating the need for additional tools or interfaces.

27:28Two updates in end-user computing. Amazon Workspace's Personal now allows you to route streaming traffic privately between your VPC and Workspace's virtual desktops using AWS PrivateLink without the data ever traversing the public internet. Definitely like this if you're someone that's got compliance requirements and you also have a Workspace virtual desktop requirement. this is definitely something that I suggest checking out. We also announced Research and Engineering Studio on AWS version 2025.06, which introduces significant improvements to instance bootstrapping, security configurations, and logging capabilities.

28:17Up next, we have one update in the internet of things. We've announced the general availability of managed integrations. And this is a feature within AWS IoT device management. And for those who are doing IoT things and you haven't checked out AWS IoT, I definitely highly recommend looking into it, especially this feature within device management. I know a lot of times customers who have IoT devices and if you need to manage a fleet of these, there comes a point at a certain scale where you either are building this out yourself of having to manage all those device software updates, security vulnerabilities, patches across all of the devices wherever they are roaming around in the world.

29:06And device management is a feature that just makes that a lot simplified all in one place. And so now, Now, what's even more exciting is that additional integrations, these could be things like hub SDKs from ZigBee, Z-Wave. This is Simon's saying, like, if you're in the know, you know what these are. I'm not an IoT person, so I don't know what these are, but you probably know if you're listening. These are some things and Wi-Fi protocols that are now just integrated within IoT device management that are going to make it a lot easier for you. Now we've got some updates in management and governance.

29:48AWS Config Rules adds classification information from AWS Control Tower Catalog. This makes it easier for you to identify how Config Rules map to different compliance frameworks, such as CIS, V8, FedRAMP, and NIST. Config Rules, they help you automatically evaluate your AWS resource configurations for desired settings. and you don't necessarily have to use config for compliance. I know a lot of people usually think about config when they're thinking about compliance requirements, but config could also be used if you have your own company's requirements for how the resources within AWS should be utilized and governed.