AWS 2025 Announcements Wishlist

Before heading to AWS re:Invent this year, my team and I put together a wishlist of anticipated announcements—features and services we hoped for. It was exciting to see some of them become reality! In this post, I'll walk through our wishlist for re:Invent 2024 but now for 2025 and discuss their real-world use cases.

Amazon Q .NET (Announced)

The previous .NET Modernization tool, AWS Porting Assistant, relied primarily on predefined rules, limiting its ability to modernize anything outside those constraints. With the introduction of Generative AI, Amazon Q .NET delivered more accurate results in one of our modernization projects.  

You can read more about our experience in the linked blog post.

Amazon Q Cobol (Announced)

Mainframe/COBOL-based systems are still widely used and are among the few that require specialized IBM hardware, with no direct support from cloud providers. The only way to migrate these workloads is through modernization.  

AWS previously acquired BluAge, a company specializing in Mainframe modernization, but its rule-based approach came with high costs. We explored it in the past but couldn’t proceed with real customers due to its limitations. In contrast, Amazon Q Cobol leverages Generative AI, which we expect to offer better accuracy and a more budget-friendly solution.

Autonomic Computing

IBM introduced the concept of Autonomic Computing in 2001, and we continue to see advancements toward realizing this vision.

Ref: IBM Autonomic Computing and Solution Installation David Cole IBM AC Customer and Partner Programs

The levels from 1 to 5 define the maturity of a computing system. Currently, we mostly experience levels 2 and 3. Despite more than two decades of progress, the advancement has been slow. AWS is in a position to introduce additional functionality within existing services to help advance towards level 3 (Predictive) and level 4 (Adaptive). In particular, functions addressing level 4 could be implemented.

Here are some of the functions that come to mind for potential inclusion in existing AWS services:

• Scalability use case: The system learns the business's peak traffic or computing periods based on historical data and adapts its scalability accordingly. This can be influenced by external factors such as:

  • Weather conditions

  • Political climate

  • Sociocultural events

  • Public events or news that affect the business

  Security use case: Blocks the addition of annotations to Kubernetes in response to the latest NGINX vulnerability.

Carbon Calculator

AWS has several future plans to become more carbon neutral, including optimizing data center energy use. Currently, AWS data centers are more than 4 times more efficient than traditional on-premises data centers.

Some AWS services are also known for their energy efficiency, such as ARM-based Graviton instances, which offer over 60% better energy efficiency.

The Carbon Calculator could be a tool that calculates customer workloads based on the AWS services they are using and provides additional discounts based on the Carbon Index (CI) calculated by AWS. This could encourage AWS customers to shift more towards carbon-neutral services.

Cloudfront Built-in Functions

CDN (Content Delivery Network) is a large market with several players. When examining the maturity of market products, we see various built-in features, such as:

• Image resizing

• Image framing

• Image watermarking

For a full list of features from a major competitor, Akamai, you can view their offerings here.

In CloudFront, these features are not built-in. However, some Marketplace products can address these needs, or you can create custom Lambda functions. While possible, this requires additional effort and is not ideal for teams with limited AWS expertise.

To better meet advanced CDN requirements, we would love to see these functions offered as built-in features in CloudFront.

CQRS/Event Sourcing

ScaleCube, as defined, has three dimensions, and AWS/Kubernetes currently addresses only one of those: Horizontal Scaling. The other two dimensions are closely tied to software architecture, but could AWS do something to support them?

One dimension is functional decomposition, which opens the door to distributed architectures but also introduces challenges in managing distributed transactions.

One way to address this is by replacing strict consistency with eventual consistency. As we know, many distributed systems that require scaling, like AWS, rely on eventual consistency. However, AWS still lacks services or features that explicitly support this. This may be a controversial topic, as it heavily depends on software architecture, but AWS could still be in a position to develop open-source libraries that help developers leverage eventual consistency. These libraries could be pre-built on selected event store and event stream services.

DB Saving Plans

I know many of you are waiting for this, but AWS customers who are benefiting from Savings Plans on Compute are also looking for similar options for RDS.

EVS (Announced during re:Invent)

VMware customers are seeking alternatives following the recent licensing changes, prompting the industry to explore other options. AWS has also announced the EVS (Preview) service to provide VMware customers with an alternative path to AWS. Tech community is now waiting the GA(Generally Available) version of this service.

At Kloia, our focus is on modernization, and our "VMware Exit" strategy is based on KubeVirt, which we'll be discussing next:

Kubevirt

We have been following KubeVirt since its release, and our evaluation of it was completed quickly. At Kloia, we are currently migrating workloads from VMware directly to KubeVirt on AWS.

Here’s a brief value proposition for KubeVirt: Running VMs within pods like containers, which merges the distinct worlds of VMs and containers under Kubernetes, simplifies infrastructure management. Sounds interesting, right? :)

Although not officially announced during re:Invent, AWS was the first cloud provider to announce support for OpenShift Virtualization (KubeVirt).

The typical architecture with KubeVirt looks like this:

Ref: OpenShift Virtualization on Red Hat OpenShift Service on AWS (ROSA)

FinOps

With the release of the latest FinOps framework by the FinOps Foundation, organizational requirements and practices around FinOps have evolved.

AWS already offers cost-related tools such as tagging, Cost Explorer, and Trusted Advisor. While these support FinOps efforts to some extent, we believe AWS could develop an end-to-end dashboard that fully supports all processes and requirements defined in the FinOps framework.

Managed Backstage

Platform Engineering has been evolving rapidly, addressing gaps that organizations still face in the DevOps movement.

IDP (Internal Developer Platform) is a key tool in platform engineering, providing delivery teams (development teams) with all the necessary functions to consume the platform.

Backstage, developed by the Spotify team and open-sourced to the community, is one of the most popular IDPs.

During re:Invent 2024, Spotify independently announced a managed Backstage service. We hope AWS will recognize the demand from platform engineering teams and bring this managed service to AWS as well.

Multi-Control Tower Organizations

As a constraint, an organization can only have one Control Tower on AWS, which limits the ability to govern multiple financial accounts under a single account. Additionally, services like SSO, which are managed at the organization level, highlight the need for separating the organizational management account from the financial management account.

I understand this has been a long-standing request from AWS, and given its impact on financial and billing processes, it requires careful consideration. However, we are eager to see this functionality implemented soon on AWS.

RAG

Retrieval-Augmented Generation (RAG) is a well-known technique used in GenAI projects. But what if AWS were to introduce RAG-specific services, such as:

• Amazon Bedrock RAG Accelerator: A fully managed service to streamlineretrieval-augmented generation pipelines with native support forAmazon Kendra, OpenSearch, and vector DBs.
• AWS RAG Agent for Bedrock:A serverless function that automatically manages retrieval, embedding storage, and grounding to improve response accuracy.
• RAG-Aware API Gateway: A managed API that can automatically handle retrieval latency, caching, and chunking strategies for LLM applications.

SLM

Not everything requires an LLM (Large Language Model); we’ve found that SLMs (Small Language Models) are sufficient for many businesses. What if AWS were to introduce SLM-specific services, such as:

• Amazon SLM StudioA low-cost model hosting and fine-tuning service optimized for SLMs like Mistral or TinyLlama, with on-demand or event-driven inference.
• AWS Lambda AI ExtensionsBring SLM inference capabilities to AWS Lambda with low-memory optimizations and lightweight inference runtimes.
• Bedrock Model ShrinkerA tool that compresses and distills large models into SLMs while maintaining accuracy for specific use cases.

VectorDB Serverless

We frequently receive the same question from our customers: Which Vector DB do you recommend?

Here are some Vector DB-related features that AWS could develop:

• Amazon Aurora Vector EditionAdds native vector storage and search capabilities to Aurora with SQL-based embedding queries.
• Amazon DynamoDB Vector ModeA fully managed, serverless vector database optimized for fast, scalable, key-value retrieval with integrated similarity search.
• Amazon OpenSearch Serverless Vector IndexEnhancements to OpenSearch Serverless to natively support HNSW (Hierarchical Navigable Small World) indexing and approximate nearest neighbor (ANN) search.

Turkiye Localzone

AWS is a growing community and business in Türkiye, but has been waiting for a local zone to meet data residency requirements for some time. We are hopeful that AWS will announce the Türkiye Local Zone in 2025.

---------

In summary, I am hopeful that AWS will make significant announcements in 2025 that will address many of the features and improvements we've been wishing for. With the growing demand from the AWS community and the evolving needs of businesses, it's clear that there are opportunities to enhance the platform in ways that would better support both technical and organizational requirements. From more advanced services to fulfill the needs of platform engineering teams, to introducing features that cater to emerging trends like Retrieval-Augmented Generation or Small Language Models, there is much potential. Additionally, the long-awaited Local Zone in Türkiye could be a game-changer for meeting data residency requirements in the region.

Let’s see if our wishes come true in 2025! :)