Unit Cost on Shared infrastructure

Calculating unit cost on shared resources is difficult.

I first ran into this ask when I was working with customers who were building IoT solutions. Most IoT solutions use shared resources like IoTHub, Event Hubs, storage etc. However, the business builds a business proposals based on the unit cost of manufacturing and operating individual devices. Ideally, the business would like to know the cost of the shared resources as a function of unit cost.

General IoT Architecture

Devices & Ingestion Individual devices connect to ingestion point like IoTHub

Process and Storage Event data is stored and manipulated through various processes. These include Event Hubs, Functions, Notebooks, Apps, Databases, Machine Learning Model, etc.

Applications Present the final data in a business offering.

Device Type Effect on Cost

This is a framework I use to illustrate that not all devices have the same impact on shared resources.

Device Types:

• Chatty Device (in blue): sends many frequent, small messages.
• Bulk Upload Device (in orange): uploads data in large, infrequent batches.
• Complex Device (in yellow): high processing needs or complex data.

Device Type	Message Count	Message Size	Message Complexity	Storage Impact	Compute Impact
Chatty Device	High	Small	Low	High	Low
Bulk Upload Device	Low	Large	Low	High	Low
Complex Device	Low	Small	High	Low	High

This breakdown can help in distributing cloud costs more accurately among different device types in an IoT setup, guiding cost optimization strategies for each type.

Cost Types

Fixed The cost of a specific device should be shared between all devices that used the resources during a specified time period. The cost should be proportional to the device’s usage during the time period

Variable Device cost should be 1:1 between resource consumption costs.

Tracking Unit Cost

Networking Must use a proxy to capture Bytes per hr.

Storage Measure GB/hr through folder Naming convention, sampling data, or proxy

Server There is not a way to track unit costs at the server layer. Must be connected to Middleware or application.

Middleware and Applications Logging unit activity, message count, size, and time to use a proxy for Server and Storage.

Risks and Assumptions

• Solution must show cost per connected device and/or Transaction
• Solution must include all resources costs
  • Consider separating pure fixed costs independent of device/unit costs
  • Consider different accounting models for fixed/independent costs
• Cost per device should go down due to economies of scale
• Consider Standardizing measuring units as Cost Per Unit
  • Unit may be different for different companies (Seat per kilometer, device, sensor per vehicle, etc)
  • Unit can be a transaction (ticket sale, etc)
• Device are heterogeneous in message size, frequency and complexity
• Device will have different impact on Storage, Compute, etc
• Applications may use subset of entire IoT pipeline
• Shared Resources have Fixed and Variable Costs