Tech ONTAP Blogs

Unlocking GenAI with your data using Amazon FSx for NetApp ONTAP and BlueXP workload factory for AWS

robertbell
NetApp
247 Views

The true value of generative artificial intelligence (GenAI) is in leveraging your proprietary data—or “context”—to differentiate and enhance the general AI capabilities of public foundational models. This is where retrieval-augmented generation (RAG) solutions come into play. 

 

But while basic designs for RAG solutions may handle small, localized datasets, scaling data access across hybrid and on-premises environments for enterprise-grade solutions introduces significant challenges. 

 

In this post I’ll outline how Amazon FSx for NetApp ONTAP (FSx for ONTAP) and NetApp BlueXP™ workload factory on AWS (workload factory) can seamlessly unlock context-aware GenAI via simplified data management and RAG-based design. 

 

Read on, here’s what we cover:

  • What are the data challenges with GenAI?
  • How does FSx for ONTAP simplify GenAI?
  • Workload factory can help create context-aware AWS Bedrock GenAI applications
  • How to build your GenAI application data infrastructure using workload factory
  • Real-world GenAI use cases with workload factory
    • Medical chatbot for healthcare professionals
    • Factory chatbot for shopfloor management
    • GenAI chatbot for hybrid environments
  • Unlock context-aware GenAI with built-in data protection

 

What are the data challenges with GenAI? 

For any GenAI model to generate contextually accurate responses, it needs seamless access to your proprietary data. Whether you're working with a first-party or third-party AI solution, your data needs to be nearby, embedded, and retrievable.

 

However, that can be easier said than done. Especially when working with on-premises or hybrid data deployments, integrating your data sources into a RAG-based GenAI infrastructure comes with a few challenges:

  • Data integration and silos: Enterprises often store fragmented data across multiple systems, making it difficult to create a unified knowledge base. Plus, many domain-specific datasets remain on-premises due to security or compliance concerns, which can limit cloud integration.
  • Big data transfers: GenAI models need substantial data for providing high qualty responses. Re-syncing large datasets  (i.e., whenever there is a change) is costly and time-consuming, sometimes even taking days to complete and disrupting workflows.
  • Dataset duplication: GenAI applications require access to up-to-date data for accurate predictions and decision-making. Managing the data updates, especially when having to keep multiple datasets synchronized, comes with a lot of operational complexity and a good chance of introducing errors.
  • File metadata access: For compliance reasons, enterprises need full reproducibility and governance of AI pipelines. But when copying file systems to object storage, critical metadata such as file permissions, ownership, and access control associated with systems like Windows Active Directory are lost. That leads to security gaps and requires manual intervention to restore access permissions.
  • Data security: Making sure that data is secured is essential. Data needs to be stored safely and within the organization’s network boundaries when moving it to the GenAI storage environment. Also, the source data protection policies such as snapshots and backup usually aren’t applied automatically. That  means reconfiguring security protocols and data protection policies manually.

 

How does FSx for ONTAP simplify GenAI? 

Amazon FSx for NetApp ONTAP (FSx for ONTAP) is a data storage service from AWS that extends the benefits of NetApp® ONTAP® to the AWS Cloud. Designed for multi-protocol access, automatic storage scaling, and many other data management features, FSx for ONTAP offers high performance and cost optimization for enterprises whose business relies on data.

 

FSx for ONTAP acts as an intelligent data layer for GenAI that addresses the main data infrastructure challenges based on best practices. This diagram shows the Amazon GenAI stack with FSx for ONTAP:

 

image1.png

 

For the data management of your RAG-based GenAI applications on AWS, FSx for ONTAP provides a seamless path to access and manage your data with extended data mobility and infrastructure support.

  • Data mobility: By bringing your data closer to AWS AI services, such as Amazon Bedrock and Amazon Q, FSx for ONTAP minimizes data transfer time and reduces associated costs. Users of NetApp ONTAP-based systems can rely on FlexCache® to unlock low-latency access to remote datasets, and SnapMirror® to keep read-only data replicas in constant sync between environments.
  • Data infrastructure support: FSx for ONTAP preserves file system metadata, access controls, and data protection policies as data is copied to different storage formats. That means you maintain data security and compliance you had with your protocols and policies. Its built-in protection mechanisms—including Snapshot copies and scheduled replication—maintain full traceability and protect your data across environments.

For the deployment of your RAG-based GenAI applications, FSx for ONTAP exposes a REST API that allows easy data access. You can develop your custom workloads using these APIs directly, or rely on the best-practice no-dev workloads available to you by using workload factory, which we’ll take a look at next. 

 

Workload factory can help create context-aware AWS Bedrock GenAI applications 

BlueXP workload factory (workload factory) is a free-of-charge NetApp service that abstracts the FSx for ONTAP API, enabling users to build and manage RAG-based GenAI applications in AWS without any development overhead. 

 

Through its user-friendly UI, workload factory simplifies the management of data, infrastructure, and AI infrastructure configurations, so your applications are both secure and easy to align with compliance goals.

 

Here you can see the workload factory UI:

 

image2.png

 

Additionally, workload factory gives you the ability to embed your FSx for ONTAP workloads and artifacts in your custom GenAI applications and infrastructure with the workload factory APIs.

 

How to build your GenAI application data infrastructure using workload factory

Building a GenAI application with workload factory and FSx for ONTAP involves a few key steps:

 

  1. Set up your FSx for ONTAP file system: Create the FSx for ONTAP file system, and store your data sources in it by attaching the volume(s) to your clients (Linux, Windows, macOS) via standard protocols, such as NFS or SMB.
  2. Create a knowledge base: A knowledge base is a container for your data sources, allowing your GenAI model to use them for RAG. As part of your knowledge base definition, you:
    1. Select your FSx for ONTAP data sources.
    2. Configure your AWS Bedrock setup, including the embedding model with chunk size and the chat model. 
    3. Define access control via Active Directory.

Workload factory then proceeds with setting up the RAG pipeline for you, including embedding the data sources into a LanceDB vector database and automatically syncing changes to data sources. After the first sync completes, you can publish the knowledge base for access to GenAI applications.

3. Test and deploy your GenAI chatbot: The RAG-based chatbot is available in the workload factory UI and accessible for external applications via the workload factory API.

 

Check out this post for detailed step-by-step instructions on setting up a GenAI application using workload factory.

 

Real-world GenAI use cases with workload factory

Below we explore three real-world use cases to provide a practical perspective on what workload factory with FSx for ONTAP can offer for your GenAI initiatives.

Medical chatbot for healthcare professionals

Use case: Doctors need quick access to accurate medical records and drug information in real time to provide optimal patient care.

 

Using workload factory and FSx for ONTAP, healthcare organizations can integrate patient records and drug databases across regions into a centralized knowledge base. That data will be securely stored and ready to be accessed in real time thanks to data mirroring. Through access control lists (ACLs), the RAG-based solution is designed with access just to surface information in patient records. 

 

Medical professionals can interact with a chatbot, allowing them to quickly cross-reference patient records with drug databases, reducing the risk of adverse drug interactions and improving patient outcomes.

 

Here’s what the this medical use case would look like:

 

image3.png

 

Factory chatbot for shopfloor management

Use case: Factory personnel need immediate guidance when machines malfunction, and downtime leads to significant revenue loss.

 

With the GenAI-based chatbot powered by Amazon Bedrock and FSx for ONTAP, technicians can input error codes and get real-time instructions. The chatbot pulls from maintenance logs, user manuals, and schematics stored in the embedded knowledge base, guiding the staff through troubleshooting procedures.

 

Here’s what a factory chatbot use case would look like:

 

image4.png

 

The data is protected and managed via FSx for ONTAP Snapshot capabilities, cloning, and replication, all of which reduce downtime and improve operational efficiency as technicians receive immediate, context-aware guidance without the need for expert intervention.

GenAI chatbot for hybrid environments

Use case: Enterprises need GenAI solutions that work seamlessly across both cloud and on-premises data environments, to maintain high-performance AI inference and low-latency access.

 

FSx for ONTAP delivers the high performance required for AI workloads, especially in hybrid deployments where data may reside both on-premises and in the cloud. Leveraging its advanced caching features, businesses can create GenAI chatbots that perform well even during cloud-bursting scenarios, where high-demand workloads are dynamically shifted between cloud and local resources. 

 

Here’s what the hybrid use case for a GenAI chatbot would look like:

 

image5.png

 

The RAG-based solution provides the chatbot with access to the most up-to-date information from both environments. Also, FSx for ONTAP retains metadata and enforces security policies across different data sources, keeping sensitive information protected, regardless of where it’s accessed.

 

Unlock context-aware GenAI with built-in data protection

With FSx for ONTAP, you can address critical data challenges like managing large datasets, retaining metadata, and ensuring data security; and, with workload factory, you can leverage no-code workloads to develop RAG-based GenAI applications that have access to your FSx for ONTAP sources. 

 

By combining these two technologies, businesses can accelerate adoption of GenAI applications that deliver real-time, contextually accurate insights, without the overhead of managing complex data environments.

 

Ready to take the next step? Visit the BlueXP workload factory homepage or get started now with a free account.

Public