I'm looking for insights into the costs associated with Azure Functions, virtual machines, and App Services, specifically for an internal application that will initially support around 5,000 users. This app needs to store data permanently, and it will experience high usage during peak hours (12 hours a day, 5 days a week). Additionally, there's potential for scaling up to 50,000 users if we include clients and affiliates in the future. Are there any metrics or breakdowns available based on traffic that could guide my decision?
5 Answers
I've been diving into Azure cost optimization for apps like yours. Azure's Cost Management can help you track expenses based on metrics like egress data (around $0.087/GB for the first 10TB/month) and compute hours. However, it’s a bit tricky since the traffic view isn't straightforward out of the box. You might end up piecing things together with Monitor metrics or API Management for better granularity, particularly when scaling, as spikes can significantly increase costs.
Exactly! Using ADLS Gen2 for storage can be a budget-friendly option, especially at that user volume.
For something like yours, I'd recommend considering Azure Container Apps. It's relatively affordable at about $40 a month, which is great for starting out.
It's important to understand your application's architecture and efficiency. Testing in pre-production environments with adequate load testing is essential to predict costs accurately. You won’t get a clear estimate without understanding what the app does in real-time. For instance, if the landing page creates a lot of initial calls, it could drive costs up before tapering off due to caching.
What exactly does your app do? If it’s something simple like an internal platform, the infrastructure can be less complex compared to applications requiring heavy computations like GPU inferencing, which would definitely affect your costs.
Watch out for timeouts! Processing and execution timeouts, as well as communication limits with other services, can impact performance. Virtual machines give you more control over these factors, while cloud native services might impose limits that aren’t always clearly stated.
Good point! Network costs can add up quickly. Also, don't forget that retaining data forever can rack up storage fees if you don't set life cycle policies.