← AWS Podcast
AWS Podcast
#725: AWS News: FSx for Lustre introduces cost-saving storage tiers, MCP servers enhance AI development tools, and more
00:39:59
Explore FSx for Lustre's new intelligent storage tiering that delivers cost savings and unlimited sc
Loading summary
Transcript29 lines
- [00:00]
Shruti Kuparkar
This is episode 725 of the AWS podcast released on June 16, 2025. Hello everyone. Welcome back to another update show on the official AWS podcast. My name is Shruti and joining me is the ever brilliant Jillian Simon is out again this week. So it is us ladies leading the way. Gillian, how are you?
- [00:29]
Jillian Ford
Wonderful. And I love the way that you called it the official AWS podcast.
- [00:34]
Shruti Kuparkar
Official? Well, yeah. So we have so many different podcast channels here at aws and this is sort of our main channel where you can get all the updates across all the services that span different categories, different customers. But then we also have several smaller subsections or yeah, sub channels. Anyway, let's get started. We have a couple of interesting stories that we want to highlight right at the outset. The first one is the introduction of a new storage class, Amazon FSX for Lustre Intelligent Tiering. This is a new storage class that delivers virtually unlimited scalability and it's the only fully elastic Lustre file storage and the lowest cost file storage Lustre file storage in the cloud. So Jimin, tell us a little bit about it. Like what is this new intelligent tiering? What does it cost? What is it good for?
- [01:41]
Jillian Ford
I love this. I mean this is literally like the set it and forget it where if you want AWS to figure out the most cost optimized way to be able to store your stuff, FSX for Lustre is the one, the intelligent tiering class of course. And the price start is less than.0005 per gigabyte per month it auto and it's going to have optional SSD cache, improve performance for your most latency sensitive workloads. And I just realized I forgot to talk more about the different storage classes within FSX for Lustre. I got just super excited. So there's actually three. There is the frequent access, infrequent access and archive. And so FSX is going to be intelligently put your data in one of those three options.
- [02:37]
Shruti Kuparkar
Yeah. So this helps you manage costs based on your access patterns. So if you have cold data, like really really cold data, maybe it sends it to archive the one that's frequently accessed, it is categorized as frequent access. And that allows you to to optimize costs. Because of that sort of a tiering, it sounds like it delivers up to 34% better price performance than on premises HDD file systems, which is huge. And I do know that like FSX for Lustre is very very popular when it comes to our model training pipelines and so on and so forth. And I wonder how, how this will impact those teams as well, because a lot of that data when the training is actually happening is very much frequent access. But then once you're done with the training, there could be times where it's just sitting there and unfortunately also costing our customers. But now FSX will move it once it's not being accessed as frequently, just to an archive stage. Okay, the next one we have is MCP servers for Amazon ecs, Amazon EKS and AWS Serverless. So first of all, let's just talk about what MCP is. MCP is the Model Context Protocol, which is essentially this protocol that talks about how large language models talk to each other, talk to other systems. And this is the launch of McP servers for ECS, EKS and AWS serverless, which means that intelligent these LLMs can now talk to these services and get an insight of what's going on under the hood, change things, launch new nodes if required, get observability or insights into what is happening, so on and so forth. So they really help AI development assistants with capabilities that enable real time contextual responses that go beyond their pre trained knowledge. Because it's like what is actually happening in my EKS environment right now and how can I control it is what these MCP servers enable. Jillian, what are some of the key capabilities here?
- [05:13]
Jillian Ford
Yeah, I think it's really cool and I'm going to give you some examples on Amazon ecs. So you could have it troubleshoot ECS deployment, you could have it get some containerized guidance. So ask it like for best practices, actually being able to generate infrastructure as code through cloudformation templates, things like automatically configure application load balancers for web traffic. So there is an AWS Labs GitHub repo. The URL is AWS Labs GitHub IO and that's the page that you can go and find more information on the Amazon ECS MCP server, to be specific.
- [05:56]
Shruti Kuparkar
Awesome. And so really what this means is that application development will be much faster with up to date AWS knowledge. Debugging and troubleshooting can be much more enhanced and deployment and configuration can be automated. Just a whole lot of benefits of course, like not to mention now you can have a natural language interface for all these complex operations. So it's exciting news for those building AI applications, especially for developers focused on developers.
- [06:33]
Jillian Ford
Yeah, it really is. And one thing I'll call out. So in this repo that I mentioned, AWS Labs GitHub IO MCP I know, sorry you have to memorize that. We'll link it so you're not having to write it down. There are multiple different MCP servers that are aware that are available that Shruti was just talking about. So not only is there Amazon Eks ECS as Shruti was talking about, there's also Lambda. There's for Amazon Nova Canvas as an example. Lambda step functions Terraform so really great resource to just check out for all the fun MCP things you could do.
- [07:12]
Shruti Kuparkar
Awesome. All right, those were the two top updates for today and now maybe let's transition over to the many other interesting things that happened over the last two weeks.
- [07:26]
Jillian Ford
Yeah, let's start with the analytics so Amazon Redshift has enhanced its vacuum operations to support increased concurrency across different tables and data warehouses. This feature reduces maintenance time and improves resource utilization by allowing multiple vacuum operations to run simultaneously. Amazon Redshift now enables cluster relocation by default for RA3 provisioned clusters. This feature allows you to move a cluster to another Availability zone when resource constraints disrupt cluster operations, maintaining the same endpoint so applications can continue without modifications. Amazon Athena announces Managed Query Results this is a new feature that automatically stores, encrypts and manages the lifecycle of query results for you at no additional cost. Amazon Datazone launches an upgrade domain to Amazon SageMaker. This is a new UI within the Datazone domain. Amazon EMR now supports Read and write operations from Apache Spark jobs on AWS lake formation register tables when the job role has full table access. This capability enables data manipulation language operations including create, alter, delete, update and merge into statements on Apache Hive and Iceberg tables from within the same Apache IcePark application. Amazon OpenSearch service now supports script plugins that allow you to add new scripting languages or custom scripting functionality to OpenSearch for operations like scoring, sorting and field value transformations during search or indexing. Amazon Managed Workflows for Apache Airflow now provides the option to update environments without interrupting running tasks on supported Apache Airflow versions 2.4.3 or later. Amazon Managed Service for Apache Flink adds IPv6 endpoints.
- [09:44]
Shruti Kuparkar
All right, next we have a quick update under application integration. Amazon CloudWatch Log Insights launches Query Results Summarization and Open Search PPL enhancements to help accelerate your logs analysis. So this new log summarizer generates a natural language summary of your query results, providing you with clear, actionable insights. Interpreting log entries can be time consuming, and this way you can talk to it in natural language and get the natural language summarization that helps you Quickly identify issues and get insights from your log data. With CloudWatch Log Insights, you can interactively search and analyze your logs, Log Insights Query Language, OpenSearch Service Pipe Processing Language or PPL and OpenSearch Service Structure Query Language, which is SQL. Yeah, so this is a really exciting update. I think we've all seen AI capabilities basically penetrating everywhere and I wouldn't be surprised if this natural language summary is powered under the hood by a lot of the work we are doing on that side.
- [11:01]
Jillian Ford
I wouldn't be surprised either. And speaking of more AI updates, we have released the Strands Agents open source SDK that takes a model driven approach to building and running AI agents in just a few lines of code. Strands scales from simple to complex agents, use cases, and from local development to local deployment in production. A lot of teams I've seen at AWS are really excited about the Strands SDK. So folks such as those in Amazon Q Developer AWS Glue, those teams have already started being playing around with it and you're probably wondering if you're thinking about it. And there's a lot of frameworks, so a lot of frameworks. They'll require developers to define complex workflows for their agents and Stress tries to simplify that by making agent development in a way that you're embracing the capabilities of of the actual intelligence of the language model from chains of thought, the call being able to call tools and reflect. So now with Strands, developers just simply define a prompt, a list of tools and code to build an agent, test it locally and you can then deploy it to the cloud. And like the two strands of DNA as an example for everyone, Strands is going to connect the two core pieces of the agent together, the model and the tools and Strands then is going to plan the agent's next steps and execute tools using the advanced reasoning capabilities of the model. Pretty cool. There's already been just a lot of interest, I would say, from the open source community, so we're really excited about that. We've already seen contributions from Accenture, Anthropic, Langfuse Meta, so really cool. So this is on a GitHub repo, so remember this is open source, so you're going to find this on its own separate landing page. So the actual landing page is strands agents.com and that gives you all the documentation for Concepts tools as well as some examples that you can use to get started with whatever fun thing you're building. Amazon Lex now extends custom vocabulary support to multiple languages including Chinese, Japanese, Korean, Portuguese, Catalan, French, German and Spanish locales. AWS Healthomics now supports automatic detection of workflow description language workflows to help streamline the workflow creation process. AWS announces the Release of neuron 2.2.3 featuring enhancements across inference training capabilities and developer tools. This release moves the NXD inference library to general availability, introduces new training capabilities including context parallelism and Orpo, and add Support for PyTorch 2.6 and JAX0.5.3. AWS entity resolution introduces near real time rule based matching to enable customers to match new and existing records within seconds. Amazon Q Developer announces Support for the AgentA coding experience within the JetBrains and Visual Studio IDEs. Amazon Q developer plugin for Eclipse IDE is now generally available. Amazon Q Developer CLI now supports Anthropic's Claude Sonnet 4 and allows developers to select specific Claudsonnet models when performing tasks in the Q Developer CLI. Another good reason to test out Q Developer again. My favorite service that has a forever free tier. Amazon Q Developer now helps customers optimize costs Another reason to like Q Developer who doesn't want cost savings. This new capability helps customers find and implement opportunities for right sizing instances, purchasing savings plans and reserved instances, terminating idle resources and more all through natural language conversations. Amazon Q Developer combines recommendations from AWS Cost Optimization Hub with expert guidance to help customers accelerate their cost optimization journey and maximize savings. So you can use it to ask questions like how can I lower my AWS bill? How was that recommendation calculated? How was this instance identified as idle Questions that usually I get asked when I talk to customers and now you can get those answers right away.
- [16:02]
Shruti Kuparkar
All right, next up we have a quick update under Business Applications aws. Wickr announces the launch of Wickr File Previews. This new feature empowers organizations to protect the sensitive files and lower the risk of data loss. It allows network administrators to configure a view only mode in the Security Group section of the AWS Management Console for Wickr. And we have a few updates under Cloud Financial Management. So AWS Cost Explorer now offers a new cost comparison feature which helps customers understand cost changes between two months. The cost comparison feature automatically automatically detects significant cost changes month over month and surfaces the key factors that are driving these changes. This is really helpful for customers to get insights into their monthly cost changes and identify what the key drivers are that are driving the increase or decrease in costs so that they can optimize it either way. Cost Optimization Hub now supports savings plans and reservations preferences, which means it now allows you to configure preferred savings plans and reservation term and payment options preferences so you can see your resulting recommendations and savings potential based on your preferred commitments. AWS Invoice Summary API is now generally available. This allows you to retrieve your AWS invoice programmatically via an SDK and it's now available generally. AWS Pricing Calculator also generally available now supports discounts and purchase commitment and we have a few more updates under compute. Amazon EC2 now enables you to automatically delete underlying Amazon EBS snapshots when deregistering Amazon machine images or AMIs, allowing you to better manage your storage costs and simplify your AMI cleanup workflow. We have exciting use for those of you who run AI and HPC workloads. We have pricing and usage model updates for EC2 instances accelerated by Nvidia GPUs. First we have availability of a savings plan for P6B 200 instances. These were launched recently, but they were only available through capacity blocks for ML. Now savings plans are also available for the Blackwell GPU Powered instances, so these are the latest and greatest highest performing EC2 instances powered by Nvidia GPUs. And second, we have reduced pricing for the earlier generation instances which is P5, P5EN, P4D and P4DE, or the Hopper and the Ampere family of GPUs powered instances. These pricing reductions will apply to on demand starting June 1 and to savings plans effective after June 4. We have up to 45% reduction on P5, up to 26% reduction on P5EN and up to 33% reduction on P4D and P4DE. That is a lot of cost savings for training inference as well as any kind of high performance computing workloads. I want to call out that you should go and check out the exact price reduction based on the usage model, whether it's on demand and what type of savings plan and what type of instance. We have the details outlined in the News blog or you could also just go to the EC2 pricing page to consult. Next up AWS Compute Optimizer now supports Aurora IO optimized recommendations for Aurora DB Cluster storage. These recommendations will help you make informed decisions about adopting the Aurora IO optimized configurations to increase pricing predictability and achieve potential cost savings based on your cluster's storage patterns and usage. Okay, we have Red Hat Enterprise Linux for AWS starting with Red Hat 10, now generally available, combining Red Hat's enterprise grade Linux software with native AWS integration. AWS Deadline Cloud Monitor now supports multiple languages, allowing you to view critical job information using an expanded selection of languages.
- [20:53]
Jillian Ford
Next up is Containers. Amazon EKS Add ons now support Private CA Connector for Kubernetes Amazon EKS announces the general availability of configuration insights for Amazon EKS hybrid nodes. These new insights surface configuration issues impacting the functionality of Amazon EKS clusters with hybrid nodes and provide actionable guidance on how to remediate identified misconfigurations. Amazon EKS and Amazon EKS Distro now Support Kubernetes version 1.33. Amazon ECS has extended the length of the container exit reason message from 255 to 1024charact. Now let's talk about databases. Amazon Aurora DSQL is now generally available. This is the fastest serverless distributed SQL database with active active high availability and multi region strong consistency. Aurora DSQL enables you to build always available applications with virtually unlimited scalability, the highest availability and zero infrastructure management. It is designed to make scaling and resilience effortless for your applications and offers the fastest distributed SQL reads and writes so DSQL Active Active Distributed Architecture is designed for 99.99% single region and 99.999% multi region availability with no single point of failure and automated failure recovery. It offers multi region strong consistency which ensures all reads and writes to any regional endpoint are strongly consistent and durable. Aurora D SQL independently scales, reads, writes, compute and storage, offering the flexibility and cost efficiency to both scale up and down and to meet any workload demand without compromising performance. Super exciting stuff. Amazon RDS for postgres introduces extended support for certain minor versions. This provides you more time up to three years to upgrade to a new major version to help you meet your business requirements. During extended support. Amazon RDS will provide critical security and bug fixes for your RDS for postgres databases after the community end support for a major version. You can run your postgres databases on Amazon Res with extended support for up to three years beyond a major version's end of standard support date. Cost Optimization Hub supports recommendations for Amazon Aurora. These recommendations help you identify idle database instances and choose the optimal database instance class and storage configuration for your Aurora databases. Amazon Neptune Announces Model Context Protocol this new server makes it easier for developers and AI assistants to interact with with Amazon Neptune enabling seamless integration of graph queries into generative AI workflows. Amazon DocumentDB with MongoDB compatibility announces an updated service level agreement promising a 99.99% availability when using a multi AZ configuration. Amazon RDS for MySQL also announces extended support.
- [24:33]
Shruti Kuparkar
Okay, next up some updates under Developer Tools AWS announces the general availability of AWS CDK Toolkit Library, a Node JS library that provides programmatic access to core AWS CDK functionalities such as synthesis, deployment and destruction of stacks. This enables developers to integrate CDK operations directly into their applications, custom CLIs and automation workflows offering much better flexibility and control over infrastructure management. AWS Amplify Hosting is excited to offer customizable build instances to provide you with more memory and CPU configurations to build your applications. This allows developers to select from multiple build instances to optimize their build environment based on their application's specific needs. And I think developers can choose from three different instance types. There's the standard with 8 gigabytes of memory and 4 VCPUs large with 16 gigabytes memory and 8 VCPUs and extra large with 72 gigabytes of memory and 36 VCPUs. A quick update under End User Computing Amazon Lex now offers AWS Cloud formation support in AWS Gov Cloud Region extending infrastructure as core capabilities to government agencies and their partners and another quick update Under Gaming Amazon Gamelift Server SDKs for C, C and Go are now open source and available on the Amazon Gamelift GitHub organization. The game engine plugins and SDKs for Unreal Engine and Unity along with developer scripts have been moved under the same GitHub organization for improved accessibility.
- [26:37]
Jillian Ford
Let's talk about management and governance. The serious topic Amazon cloudwatch Now support Log's transformation for Contributor Insights. This allows you to analyze transformed or enriched logs so customers can now create Contributor insights rules on CloudWatch logs that have been transformed into JSON format to create time series that display the top end contributors, total number of unique contributors and their usage. Amazon CloudWatch Synthetics now supports the Java programming language for authoring canaries, enabling developers to write monitoring scripts using the Java 21 runtime environment. CloudWatch Database Insights announces support for Amazon Aurora Postgres Limitless Database Database Insights is a database observability solution that provides a curated experience designed for DevOps engineers, application developers and database administrators. AWS announces two enhancements to CloudTrail event enrichment, which makes it easier to categorize, search and analyze your AWS activity and expanded event size which improves visibility into API actions for more comprehensive security analysis. So with Event Enrichment you can enrich your CloudTrail management and data events with additional information relevant to your business context. AWS Resource Group APIs now come with AWS PrivateLink support. This allows you to invoke AWS Resource Group APIs from within your VPC without traversing the public Internet. Aws Resource Group APIs now supports IPv6. Amazon Managed Service for Prometheus now supports queries with time ranges up to 95 days and this is an increase from the previous 32 day limit. One quick update in Marketplace AWS Marketplace has expanded its global accessibility by introducing support for French, Spanish, Korean and Japanese languages across both the website and AWS console.
- [28:59]
Shruti Kuparkar
Next we have a quick update on our media services. Amazon Interactive Video Service or Amazon IVS Real Time Streaming now supports participation Ah Beep Amazon Interactive Video Service or Amazon IVS Real Time Streaming now supports participant Replication, allowing you to copy a participant from one IVS stage to another. With this capability, participants can appear in multiple stages simultaneously facilitating cross stage interactions. Another quick one under Migration and Transfer AWS datasync customers can now transfer data directly between storage in other clouds and Amazon S3 without needing to deploy Datasync agents. This new capability uses Datasync enhanced mode to deliver increased performance and scalability, helping customers streamline their data pipelines and accelerate migrations from other clouds to aws. And we have a few updates under Networking and Content Delivery. AWS announced the opening of a new AWS Direct Connect location within the Chief Telecom HD Data center near Taipei. By connecting your network to AWS at the new location, you gain private direct access to all public AWS regions except those in China, AWS Glove Cloud regions and AWS Local zones. AWS Network Firewall launches a new monitoring Dashboard which enhances your ability to monitor the network traffic your network traffic. The new dashboard offers really valuable insights into traffic patterns including top traffic flows, TLS Server Name indication and the HTTP host headers. AWS Network Firewall now supports configuring multiple VPC endpoints for a single firewall. This new capability gives you more options to scale your network firewall deployment across multiple Amazon Virtual private clouds or VPCs. Using a centralized security policy. Amazon VPC Route Server enhances connectivity monitoring and logging with new network metrics. These metrics allow customers to proactively monitor network health, troubleshoot network issues, and gain visibility into route propagation and peer connectivity. AWS Private Certificate Authority or AWS Private CA now supports Active Directory Child domains through the Private CA Connector for ad. This feature allows customers to get a consistent experience using AWS Private CA across parent and child AD domains. AWS Site to Site VPN a fully managed service that allows you to create a secure connection between your data center or branch office and your AWS resources using IP security tunnels is adding three new capabilities for enhanced security and ease of configuration. First, we have the AWS Secrets Manager integration which allows customers to store their pre shared keys in Secrets Manager VPN connection. API responses will redact the PSK or the pre shared keys and instead display the Secrets Manager ARN or the Amazon Resource name to provide enhanced security. Second is that there's a new API to track VPN algorithms. You can now easily track the currently negotiated Internet Key Exchange version, DH groups or encryption algorithms and integrity algorithms using the Get Active VPN Tunnel Status API. And lastly, there is now a recommended configuration parameter for the Get VPN Connection Device Sample Configuration API to help you use the best practices security configuration ENA Express A purpose built network interface powered by the SRD, the scalable reliable datagram protocol for for EC2 instances in AWS Gov cloud regions is now available. ENA Express is an ENA feature that uses the SRD protocol to improve network performance in two key ways, higher single flow bandwidth and lower tail latency for network traffic between EC2 instances.
- [34:09]
Jillian Ford
Now we're going to talk about security identity and compliance. AWS WAF now supports matching incoming requests against autonomous system numbers. By monitoring and restricting Traffic from specific ASNs, you can mitigate risks associated with malicious actors, comply with regulatory requirements, and optimize the performance and availability of your web applications. AWS Security Hub now supports NIST SP 800171 revision 2. AWS Secret Manager now enables customers to allocate and track costs for their secret usage.
- [34:51]
Shruti Kuparkar
Okay, a quick update under serverless Amazon API Gateway now supports routing rules for rest APIs using custom domain names. This allows you to dynamically route incoming requests based on HTTP header values, URL based paths, or a combination of both.
- [35:13]
Jillian Ford
And the last category in today's update show is Storage Mount point for Amazon S3 now lets you automatically mount an S3 bucket when your Amazon EC2 instance starts up. This simplifies how you define a consistent mounting configuration that automatically applies when your instance starts up and persists the mount when the instance reboots. Amazon S3 Express 1 zone now supports granular access controls with S3 access points. Amazon FSX for NetApp ONTAP now supports write back mode for Ontap Flex cache volumes. So writeback Mode is a new ontap capability that helps you achieve faster performance for your write intensive workloads that are distributed across multiple write AWS regions and on premises file systems. AWS Backup now supports Amazon Aurora D SQL AWS Backup Search now supports creating backup indexes within backup policies. This allows you to automatically create backup indexes for your Amazon S3 backups or Amazon EBS snapshots at the AWS organization level. AWS Backup adds support for SNS notifications and EventBridge events for backup indexes. AWS Backup enhances Amazon EC2 restores with custom volume configuration support. This new feature enables users to specify custom settings for all the attached Amazon EBS volumes to an EC2AMI, including volume type size, IOPS, AWS, KMS encryption keys and others directly through AWS backup. A lot of good updates Trudy, what do you think?
- [37:07]
Shruti Kuparkar
Yes, definitely a lot of exciting updates. I mean of course we covered some of the really sort of impactful ones at the start of the episode with the McP servers for EKs, ECS and serverless and also the intelligent tiering solution for FSX for Lustre. But other than those, were there any that caught your eye?
- [37:31]
Jillian Ford
I would say the Strands one I've seen just a lot of buzz just here within AWS about it for customers and just the options it has as an open source option for being able to orchestrate agents. So really excited to see what customers start building using it. What about you sri?
- [37:53]
Shruti Kuparkar
I think there were a few updates around the improved service level agreement maybe for the Amazon documentdb across, you know when using the multi availability zone and there were a few others. I just love these kind of updates where things just get better without customers having to worry about it. I mean that's the beauty of running workloads on the cloud is that there's this whole whole team who is sort of working to optimize and year over year we do better on cost optimization, we do better on our availability SLAs and yeah, so love all the innovation there.
- [38:34]
Jillian Ford
Me too.
- [38:35]
Shruti Kuparkar
Yeah it's sometimes interesting how we read out every nine in the update like two nines or three nines and every little, you know, decimal digit counts right.
- [38:49]
Jillian Ford
Like so yeah it is. I mean this Optit show actually had a lot which is now that I'm like thinking more on like your question. I mean the Aurora D SQL I liked getting cost optimization recommendations for from Q developer and just being able to ask and then get those insights because again I mean everyone's always looking for ways to reduce their costs. Like yeah, this was a jam packed one. Now that we've done it it definitely was.
- [39:17]
Shruti Kuparkar
All right, thank you for listening everyone. Those were all the updates we have. Jillian, where can listeners find you and.
- [39:24]
Jillian Ford
Get in touch with you Jillian Ford on LinkedIn.
- [39:29]
Shruti Kuparkar
Yeah. And same with me, Shruti Kuparkar on LinkedIn. But also, we would love to get feedback from you. If you go to the AWS podcast page, there is, as Gillian calls out, a big yellow button. Press it, give the feedback. We are really eager to get some insights on what you like, what you don't like, what we would like to see more of. And with that, that's the end of the show today. Until next time, keep on building.