> Markdown version of [/videos/1559-in-depth-net-azure-functions-flex-plan-isolated-mode-and-performance](https://www.wearedevelopers.com/videos/1559-in-depth-net-azure-functions-flex-plan-isolated-mode-and-performance). Every page supports `.md` or `Accept: text/markdown`. Links point to the HTML versions so they work for humans too. Agent guide: [/agents.md](https://www.wearedevelopers.com/agents.md).

---

# In-depth .NET Azure Functions: Flex plan, Isolated mode and performance

The Azure Functions in-process model is ending by .NET 10. Master the new Kubernetes-backed Flex Consumption plan to optimize cold starts and scale event-driven architectures efficiently.

- **Speakers:** [Stanislav Lebedenko](https://www.wearedevelopers.com/@stanislav-lebedenko)
- **Event:** World Congress 2025
- **Published:** August 20, 2025
- **Duration:** 29:05
- **URL:** https://www.wearedevelopers.com/videos/1559-in-depth-net-azure-functions-flex-plan-isolated-mode-and-performance

## Summary

Microsoft is fundamentally overhauling Azure Functions, transitioning away from the legacy virtual machine in-process model toward a Kubernetes-backed infrastructure known as the Flex Consumption plan. Built on "Project Legion" and sharing infrastructure with Azure Container Apps (ACA), Flex Consumption brings significant architectural changes. Ultimately, the in-process model will be deprecated in favor of the Isolated Worker mode, becoming the mandatory standard by .NET 10. This shift standardizes Azure Functions as typical .NET applications, significantly improving portability to on-premises Kubernetes or ACA environments.\n\nThe Flex Consumption plan introduces dynamic scaling capabilities and free virtual network (VNet) integration, though developers must navigate new constraints. While Microsoft advertises a maximum scale-out of 1,000 instances, default subscription limits and a strict 40-IP address pool for VNet integration require careful architectural planning—especially for long-running HTTP triggers constrained by a strict five-minute lifecycle. For event-driven architectures experiencing massive traffic spikes, Flex dramatically outperforms the classic consumption plan's scaling speed. The platform offers discrete memory allocations, where the smallest 0.5 GB instances provide the fastest volume scaling, provided the application's memory footprint is heavily optimized.\n\nTo maximize performance and cost-efficiency in this new isolated model, developers should aggressively limit .NET NuGet dependencies rather than relying on Native AOT or Ready-to-Run (R2R) compilations, which currently yield negligible cold-start benefits. Fine-tuning concurrency settings in the host configuration dictates whether it is more cost-effective to scale out massively using 0.5 GB instances or to handle denser concurrent loads on fewer 4 GB instances. Teams are strongly advised to leverage external load-testing tools rather than relying solely on Azure's built-in automatic performance optimizer, ensuring their workloads are fully prepared for the mandatory migration to the Isolated Worker mode.

**Keywords:** azure functions, flex consumption plan, isolated worker mode, kubernetes architecture, azure container apps, .NET 8 serverless, cold start optimization, vnet integration limits, memory footprint profiling, concurrency configuration, event-driven scaling, legacy in-process model, ready-to-run compilation, native aot limitations, serverless load testing

## Chapters

1. **Transitioning to the Azure Functions flex consumption plan** (00:22) — Microsoft is replacing the legacy in-process model with a Kubernetes-backed isolated worker model.
1. **Understanding the underlying architecture with Project Legion** (02:53) — Awareness of Azure container apps and Kubernetes pods helps troubleshoot deployment and scaling failures.
1. **Optimizing container sizes for platform portability** (04:04) — Smaller container footprints allow developers to package and deploy functions on any on-premise Kubernetes instance.
1. **Navigating isolated worker support and shared platform nuances** (04:58) — The flex consumption tier shares its infrastructure with Azure container apps and strictly requires the isolated worker model for .NET 8.
1. **Managing virtual network integrations and HTTP trigger timeouts** (07:06) — Limited IP address ranges and strict timeout rules cause unexpected network and processing failures during sudden traffic spikes.
1. **Analyzing memory sizes and scaling behavior in flex consumption** (10:52) — Smaller half-gigabyte instances achieve significantly faster parallel scaling compared to larger memory configurations.
1. **Comparing performance between flex and classic consumption plans** (13:40) — The legacy consumption tier running on older hardware struggles with CPU overload during rapid request spikes unlike the updated flex infrastructure.
1. **Embracing the isolated process model for future optimizations** (16:53) — Adopting the standard .NET application structure enables better feature alignment and prepares projects for upcoming framework updates.
1. **Maximizing scale speeds with minimal memory allocations** (18:01) — Provisioning half-gigabyte instances provides the fastest deployment response for handling sudden massive traffic surges.
1. **Evaluating native AOT and ReadyToRun compilation strategies** (20:27) — While native AOT lacks proper support, ReadyToRun offers negligible cold start improvements that are overshadowed by platform-level optimizations.
1. **Tuning application dependencies and host concurrency settings** (22:56) — Removing unnecessary packages and adjusting host.json concurrency thresholds balances processing performance with infrastructure costs.
1. **Automating performance testing and containerizing Azure Functions** (25:48) — External load testing tools provide more reliable metrics than built-in optimizers when preparing function code for Kubernetes migration.
1. **Monitoring SDK updates and mitigating platform limits** (27:16) — Tracking GitHub issues helps avoid broken package releases while combining hosting plans circumvents restrictive subscription compute caps.

## Related Moments

- [Azure functions hosting plans and migration challenges](https://www.wearedevelopers.com/videos/100207-in-depth-net-azure-functions-isolated-mode-performance-and-durable-ai-agents) (from "In-depth .NET Azure Functions: Isolated mode, performance and durable AI agents")
- [Adopting the .NET isolated worker model for functions](https://www.wearedevelopers.com/videos/100207-in-depth-net-azure-functions-isolated-mode-performance-and-durable-ai-agents) (from "In-depth .NET Azure Functions: Isolated mode, performance and durable AI agents")
- [Transitioning from pure code functions to serverless containers](https://www.wearedevelopers.com/videos/1008-serverless-java-in-action-cloud-agnostic-design-patterns-and-tips) (from "Serverless Java in Action: Cloud Agnostic Design Patterns and Tips")
- [Comparing serverless functions against traditional monolithic architectures](https://www.wearedevelopers.com/videos/590-building-your-way-to-a-serverless-powered-iot-buzzwire-game) (from "Building your way to a serverless powered IOT Buzzwire game")
- [Running microservices on Azure Container Apps](https://www.wearedevelopers.com/videos/440-back-end-to-the-future-embracing-the-continuous-evolution-of-infrastructure-and-code) (from "Back(end) to the Future: Embracing the continuous Evolution of Infrastructure and Code")
- [Resolving resource scaling challenges with dynamic serverless deployments](https://www.wearedevelopers.com/videos/1008-serverless-java-in-action-cloud-agnostic-design-patterns-and-tips) (from "Serverless Java in Action: Cloud Agnostic Design Patterns and Tips")

## Related Articles

- [What Are The Top Skills Required For Azure Developers?](https://www.wearedevelopers.com/magazine/77-what-are-the-top-skills-required-for-azure-developers)
- [Building AI Solutions with Rust and Docker](https://www.wearedevelopers.com/magazine/494-building-ai-solutions-with-rust-and-docker)
- [Dev Digest 132 - Binging WADFlix?](https://www.wearedevelopers.com/magazine/473-dev-digest-132-binging-wadflix)
- [Dev Digest 120 - Apple and peers](https://www.wearedevelopers.com/magazine/455-dev-digest-120-apple-and-peers)

## Related Jobs

- [Staff Software Engineer, Compute Platform](https://www.wearedevelopers.com/jobs/ext/2615501-staff-software-engineer-compute-platform) at **GitHub**
- [Staff Software Engineer, Performance Tooling](https://www.wearedevelopers.com/jobs/ext/3583824-staff-software-engineer-performance-tooling) at **ARM**
- [GPU Cluster Engineer, Networking](https://www.wearedevelopers.com/jobs/48410-gpu-cluster-engineer-networking) at **Sciforium**
- [Principal Forward Deployed Engineer, Performance](https://www.wearedevelopers.com/jobs/48480-principal-forward-deployed-engineer-performance) at **Arm**
- [Senior AI Serving Engineer, Backend](https://www.wearedevelopers.com/jobs/48414-senior-ai-serving-engineer-backend) at **Sciforium**
- [Staff SoC Performance Architect - Next-Generation Server Platforms](https://www.wearedevelopers.com/jobs/ext/1916643-staff-soc-performance-architect-next-generation-server-platforms) at **ARM**