Skip to main content

Cloud Modernization for AI: Serverless and Containerization (Pill 3 of 5 / Cloud Pills)

AI is reshaping the way we build and run businesses in all industries. One challenge with AI models is scaling. Traditional infrastructure is not built for the dynamic scaling required by AI models during training or operation.

 


To optimise cost and reduce operational overhead, a modern approach combining serverless and microservices to provide a flexible, scalable, and efficient workload layer is required. Microsoft Azure enables these two mechanisms through Azure Functions and Azure Kubernetes Services.

Serveless is required for AI deployments, especially because of unpredictable demands. The capability of running a function triggered by an AI agent in response to an event without the overhead of deployments is crucial for multi-agent AI solutions. Serverless is needed for real-time image recognition, language translation and dynamic execution of payloads triggered by APIs, data streams and IoT devices.

Another advantage of a serverless approach is agility and the ability to deploy AI models with fewer infrastructure concerns, increasing the iteration cycles. It enables developers and data scientists to deploy their solutions more easily and quickly. If 10 years ago, deployment in a Development environment every 1-2 days was normal, nowadays, 20-30 deployments per day are part of the way of working using service infrastructure.

Azure Functions provides the agility, scaling and flexibility required to build such a solution on top of the Microsoft ecosystem. It combines the serverless benefits with strong integration with other services and the security layer of Microsoft Azure.

Running AI models in a serverless approach can be challenging and has drawbacks. For instance, a 'cold start ', a delay in the execution of a function due to its initialisation, can impact the AI system's responsiveness. Resource limitations, especially regarding GPU and CPU, stateful processing, limited or no GPU support, and cost optimisation at scale, are other potential challenges.

Kubernetes provides a managed environment for a container-based approach, enabling AI models to run and scale successfully. Azure Kubernetes Services (AKS) can also successfully run AI workloads, providing high computation resources in a cost-efficient environment.

A chatbot with NLP capabilities deployed across AKS can scale across multiple nodes of the same cluster in a high-availability and low-latency environment.

Compared with a native serverless approach, Kubernetes and AKS can successfully manage and orchestrate GPU and CPU workloads. With native support for GPU nodes, an AI model can be trained on GPU nodes without additional infrastructure complexity. Kubernetes allow organisations to share the GPU nodes of their backend across multiple models, reducing processing time while keeping GPU costs low.

Azure Functions and AKS run in a native cloud environment. There are scenarios where multi-cloud and hybrid deployments are required. For additional flexibility, Azure Arc extends AKS capabilities, allowing organisations to run AI models in non-cloud environments or edge locations. This is a perfect solution for organisations requiring data sovereignty and localised AI models. The healthcare industry is a good example, where processing patients' data might require AI models that run on-premise environments to comply with local laws.

 

The combination of serverless and Kubernetes in a modern computation approach enables organisations to run their AI workloads cost-effectively, event-driven, and scalable. Azure Functions and AKS are two services that respond to AI deployment needs with advanced orchestration capabilities. With the strong support of containerisation for AKS and Azure Functions, Azure Arc brings cloud capabilities to locations where Azure Region or cloud cannot be used.

The combination of Azure Arc, Azure Kubernetes, and Azure Functions enables customers to build, run, and manage AI applications in a multi-cloud and hybrid approach while keeping the technical stack and debt under control.  

Comments

Popular posts from this blog

How to audit an Azure Cosmos DB

In this post, we will talk about how we can audit an Azure Cosmos DB database. Before jumping into the problem let us define the business requirement: As an Administrator I want to be able to audit all changes that were done to specific collection inside my Azure Cosmos DB. The requirement is simple, but can be a little tricky to implement fully. First of all when you are using Azure Cosmos DB or any other storage solution there are 99% odds that you’ll have more than one system that writes data to it. This means that you have or not have control on the systems that are doing any create/update/delete operations. Solution 1: Diagnostic Logs Cosmos DB allows us activate diagnostics logs and stream the output a storage account for achieving to other systems like Event Hub or Log Analytics. This would allow us to have information related to who, when, what, response code and how the access operation to our Cosmos DB was done. Beside this there is a field that specifies what was th...

Can User Technical Profile Influence AI Architecture Decisions?

In the last few days, I ran a small experiment on Devin/Windsurf/Cascade, BMAD, and LLM-based architecture decisions. The initial concern was simple: if an AI coding assistant knows something about the user, for example, that the user is a Java developer or a .NET developer, can this information influence the architecture and technology stack that the AI will propose? This question is important because many companies are starting to use AI tools not only for coding but also for product discovery, architecture, documentation, and technical decision-making. If the AI is influenced by the developer's personal profile, the result may not be fully neutral. It may look like a business or architecture decision, but in reality, it may be partially shaped by the context of the person using the tool. The test was done using BMAD with Devin/Windsurf/Cascade. The scenario was a unified field service operations platform. The task was to let BMAD refine the product, define the architecture and s...

ADO.NET provider with invariant name 'System.Data.SqlClient' could not be loaded

Today blog post will be started with the following error when running DB tests on the CI machine: threw exception: System.InvalidOperationException: The Entity Framework provider type 'System.Data.Entity.SqlServer.SqlProviderServices, EntityFramework.SqlServer' registered in the application config file for the ADO.NET provider with invariant name 'System.Data.SqlClient' could not be loaded. Make sure that the assembly-qualified name is used and that the assembly is available to the running application. See http://go.microsoft.com/fwlink/?LinkId=260882 for more information. at System.Data.Entity.Infrastructure.DependencyResolution.ProviderServicesFactory.GetInstance(String providerTypeName, String providerInvariantName) This error happened only on the Continuous Integration machine. On the devs machines, everything has fine. The classic problem – on my machine it’s working. The CI has the following configuration: TeamCity .NET 4.51 EF 6.0.2 VS2013 It see...