Azure Functions: Hosting Plans

Azure Functions offers three hosting plans, each with different scaling behaviour, performance characteristics, and pricing models. Choosing the right plan is critical — it affects cold start latency, maximum execution time, networking capabilities, and cost. This lesson covers the Consumption, Premium, and Dedicated (App Service) plans in detail.

---

Plan Overview

Feature	Consumption	Premium (Elastic Premium)	Dedicated (App Service)
Scaling	Event-driven, 0 to 200 instances	Event-driven, 1+ pre-warmed instances	Manual/autoscale rules
Scale to zero	Yes	No (minimum 1 instance)	No
Cold start	Yes (1–10 seconds)	Minimal (pre-warmed instances)	None (always running)
Max execution	10 minutes	Unlimited	Unlimited
Max instances	200	100	10–30 (varies by SKU)
VNet integration	No	Yes	Yes
Pricing	Per execution + GB-s	Per pre-warmed instance + burst	App Service plan cost
SLA	99.95%	99.95%	99.95%

---

Consumption Plan

The Consumption plan is the default and most cost-effective option for intermittent workloads. Azure manages all infrastructure and scaling decisions.

How It Works

1. An event arrives (HTTP request, queue message, timer)
2. The scale controller evaluates the event backlog
3. New instances are added as needed (up to 200)
4. When events stop, instances are deallocated back to zero

Pricing

You pay for two things:

• Executions — $0.20 per million executions (first 1 million free)
• Resource consumption — $0.000016 per GB-s (first 400,000 GB-s free)

Example: 5 million executions/month, 128 MB memory, 500ms average duration

Execution cost:    (5M - 1M free) × $0.20/M = $0.80
Resource cost:     5M × 0.125 GB × 0.5s = 312,500 GB-s
                   (312,500 - 400,000 free) = $0.00 (within free grant)

Total: $0.80/month

Limitations

• Execution timeout — 10 minutes maximum (default 5 minutes)
• Cold starts — first invocation after idle incurs startup latency (1–10 seconds)
• No VNet integration — cannot access resources in a virtual network
• Limited CPU/memory — 1.5 GB memory maximum per instance
• Storage — requires an Azure Storage account for internal state management
• Outbound IPs — shared and unpredictable (cannot whitelist)

Best For

• Prototypes and MVPs
• Low-traffic APIs
• Scheduled tasks with infrequent execution
• Event processing with variable load
• Cost-sensitive workloads

---

Premium Plan (Elastic Premium)

The Premium plan eliminates cold starts while retaining event-driven scaling. It is designed for production workloads that need consistent performance and enterprise networking.

How It Works

1. You configure a minimum number of pre-warmed instances (at least 1)
2. Pre-warmed instances are always running and ready to handle requests instantly
3. When demand exceeds the pre-warmed count, additional instances are added automatically
4. Burst instances scale back when demand drops, but pre-warmed instances remain

Pre-warmed instances:  ████████  (always running — no cold start)
Burst instances:       ░░░░░░░░  (added on demand — may have slight delay)

Pricing

SKU	vCPUs	Memory	Price (approx.)
EP1	1	3.5 GB	~$220/month
EP2	2	7 GB	~$440/month
EP3	4	14 GB	~$880/month

You pay for:

• Pre-warmed instances — always billed (per-second)
• Burst instances — billed per-second only when active

Key Features

• No cold starts — pre-warmed instances handle requests immediately
• Unlimited execution duration — no timeout limit
• VNet integration — access resources in Azure Virtual Networks
• Private endpoints — expose functions only within a VNet
• More resources — up to 14 GB memory and 4 vCPUs per instance
• Predictable outbound IPs — can be whitelisted by downstream services

Best For