How One Routing Change Could Cut an IoT Bill by 85%

Consider an illustrative manufacturing fleet of 8,000 vibration sensors, each sending 120 readings per second. Assume the Azure bill runs $22,000 a month for IoT processing, while fewer than 3% of readings require an alert within 30 seconds. Most telemetry passes through real-time processing even though it can tolerate a delay.

If routing those messages into hot, warm, and cold paths brought the processing bill to $3,100, the reduction would be about 85%. This hypothetical example shows how routing can affect cost. Actual savings depend on message volume, processing requirements, service configuration, and retention needs.

The Single-Pipeline Trap

IoT teams often start with one pipeline: devices send telemetry, a stream processor evaluates it, and results go to a dashboard and a database. This approach can serve a proof of concept with 50 devices and a few dashboards.

As the fleet grows, urgent alerts and routine records compete for processing capacity. A safety-critical temperature reading may share a queue with a compliance metric that will not be reviewed for 30 days. Processing both in real time adds cost and can delay the messages that require immediate action.

Three Paths, Three Purposes

Route telemetry according to how quickly someone or something needs to respond. Each path should have a processing timeline and cost profile that fit its purpose.

Hot path: act within seconds. This path supports alerts, automated commands, and live dashboard updates. Reserve it for events that require an immediate response. As a starting guideline, ask whether a system or person will act on the data within 60 seconds.

Warm path: analyze within minutes. This path supports anomaly detection, rolling averages, and trend analysis. It processes information soon enough to support operational decisions while allowing minutes of delay.

Cold path: store for later. This path supports archival, compliance reports, and historical analysis. Data goes to lower-cost storage and is processed in batches when needed.

The share of telemetry assigned to each path depends on the product and its response requirements. Keep the hot path focused on urgent events, and use warm and cold processing where the application can tolerate a longer delay.

What This Looks Like for One Connected Product

Consider a commercial refrigeration unit. The same unit can produce information that supports an immediate response, a service decision, and a historical report. Each use calls for a different processing timeline.

Data or eventBusiness needPath
Temperature exceeds a critical thresholdNotify an operator quickly enough to protect inventoryHot
Temperature gradually rises over several hoursIdentify a developing issue and plan serviceWarm
Historical temperature recordsSupport reporting and longer-term analysisCold

The paths do not have to be mutually exclusive. A critical reading may trigger an immediate alert and also be retained for later analysis. The decision is how quickly each use needs to be served.

The Unexpected Benefit

Separating routine archival traffic from urgent messages can also improve alert performance. When congestion is delaying alerts, moving less urgent processing to other paths can free capacity and shorten response times.

Measure alert latency and data completeness alongside spend to confirm that a routing change delivers the intended result. The revised architecture must continue to support the alerts, reporting, and historical records that customers rely on.

Turn Telemetry Into Action With MeshInsights

Routing determines how quickly data is processed. MeshInsights extends that workflow with AI agents that interpret connected-product data, diagnose issues, and recommend or initiate a service response within customer-defined controls.

By automating recurring decisions, MeshInsights helps service teams support more connected equipment without increasing expert workload at the same rate. Experts remain involved when cases require additional judgment, while routine issues move through a more consistent response process.

When to Revisit Your Data Architecture

Review routing as your fleet grows and customer requirements change. These signs can help identify where to investigate:

  • Cloud processing costs are growing faster than the connected fleet.
  • Alerts slow down during spikes in routine telemetry.
  • Data collected for occasional reports receives the same processing as urgent events.
  • The architecture still follows the original proof of concept, despite changes in scale and customer needs.

Trace representative message types from ingestion to the customer action or report they support. Compare the required response time with the processing each message receives to determine whether routing contributes to the problem.

Three Questions to Ask Your Engineering Team

Start the review by asking your engineering team how processing decisions support customer needs. Use these questions to establish the requirements and identify potential changes:

  1. Which events require action within seconds, and what happens if we respond late? Establish the response time each use case needs.
  2. How much of our processing spend supports those urgent events? Identify the costs of immediate action, enrichment, and retention.
  3. What information can we process less frequently while still meeting customer and operational needs? Select a limited routing change and define how to measure its cost and performance.

Use the answers to agree on a first change across product, engineering, and finance. Measure the results before extending the approach to other message types.

Key Takeaways

  • Route telemetry by urgency. Match processing timelines to the action or decision each message supports.
  • Use the 60-second guideline. Ask whether a system or person needs to act within that window when assessing hot-path requirements.
  • Measure cost and performance together. Check alert latency and data completeness as you reduce processing spend.
  • Validate savings against your workload. The illustrative 85% reduction shows the potential impact of routing. Your own costs and response requirements determine the opportunity.

Dive Deeper

Read Hot, Warm, and Cold: Designing Data Paths That Scale Without Breaking the Bank for the technical implementation. The article covers Azure service mappings, Stream Analytics queries, routing configuration, and a cost comparison.

Kyle Burns is a Practice Lead, Technical Principal at Mesh Systems.

  • Solutions
  • Industries
  • Company
  • Resources