Analytics & ML · Study deck

Edge Processing: Aggregation Decisions

A threshold that saves bandwidth can also erase a short alarm.

Data Dora is your guide for this deck.

edgepatternsprocessing

Data Dora, the module guide, in a scene from this chapter.

After studying this chapter

Learning objectives

You will be able to:

Explain: At that scale, cloud compute costs grew to ~$500/month while edge gateways were already provisioned with spare capacity, making the per-sensor cost of edge processing lower than cloud.
Explain: Cost savings need enough volume: the chapter's decision framework separates small deployments, medium deployments, and large deployments above 10,000 sensors or 10 GB/month.
Explain: Hybrid architectures run simple detection at edge (anomaly flags, threshold checks) while cloud handles deep analysis (root cause diagnosis, long-term forecasting).
Explain: An inference step sends anomaly evidence, while Store-Forward keeps buffered readings until the network returns.

Major section

Checkpoint: Run the Four Contracts

A threshold filter sends exceptions and discards routine readings.

An inference step sends anomaly evidence, while Store-Forward keeps buffered readings until the network returns.

Farm edge gateway aligns heterogeneous sensors, combines redundant readings, buffers an outage, and writes a lineage-preserving cloud record.

Major section

Edge ML vs Cloud ML Inference

Hybrid architectures run simple detection at edge (anomaly flags, threshold checks) while cloud handles deep analysis (root cause diagnosis, long-term forecasting).

A factory safety system needs 20 ms edge response, but weekly predictive maintenance reports can use cloud-trained models updated monthly.

Major section

Batch vs Streaming at Edge

Option A (Batch Processing): Collect sensor data locally for 1-60 minutes, process in batches, upload aggregated results.

Option B (Real-Time Streaming): Process each sensor reading immediately as it arrives with sub-second latency.
A smart meter can batch hourly readings, but a pipeline pressure sensor must stream in real-time to detect ruptures within seconds.

Major section

Edge Does Not Always Save Cost

Edge aggregation reduces cloud traffic by 90%: 36 MB/month.

Software updates: Edge devices require OTA update infrastructure.
At that scale, cloud compute costs grew to ~$500/month while edge gateways were already provisioned with spare capacity, making the per-sensor cost of edge processing lower than cloud.

Major section

Checkpoint: Price the Trade-off

Edge ML is attractive for 10-50 ms decisions, while cloud ML often pays a 200-500 ms round trip.

Cost savings need enough volume: the chapter's decision framework separates small deployments, medium deployments, and large deployments above 10,000 sensors or 10 GB/month.

Major section

Danfoss Refrigeration Edge

Each store has 12 display cases with 3 sensors each (temperature, defrost status, compressor current), reporting every 30 seconds.

40,000 stores x 12 cases x 3 sensors = 1,440,000 sensors.
Key Insight:: The three sensor types in the same deployment use three different edge patterns.
Pattern selection is per-sensor, not per-deployment.

Major section

Checkpoint: Validate the Fleet Mix

The Danfoss case starts with 40,000 stores, 12 display cases per store, and 3 sensors per case.

The raw daily total is 33.2 GB, and the mixed edge plan reduces it to 1.49 GB.

Deck summary

Key takeaways

A threshold filter sends exceptions and discards routine readings.

Hybrid architectures run simple detection at edge (anomaly flags, threshold checks) while cloud handles deep analysis (root cause diagnosis, long-term forecasting).
Option A (Batch Processing): Collect sensor data locally for 1-60 minutes, process in batches, upload aggregated results.
Edge aggregation reduces cloud traffic by 90%: 36 MB/month.
Edge ML is attractive for 10-50 ms decisions, while cloud ML often pays a 200-500 ms round trip.

Retrieval practice

Recall check 1 of 2

Data Dora says: answer from memory, then check your reasoning.

Q1A gateway uses the chapter’s aggregation function on a window of readings. What should it send upstream?

AA summary containing statistics and window context

BA separate upload for each original reading

CAn inference label replacing the window statistics

DA buffered batch awaiting restored connectivity

Show answer

Answer: A The example returns minimum, maximum, average, deviation, count, and window end.

Q2An agricultural gateway buffers readings during a link outage. In the example, when should it clear the buffer?

AAfter a network connection becomes available

BAfter copying readings into an upload batch

CAfter accepting the next sensor reading

DAfter the batch upload reports success

Show answer

Answer: D The buffer is cleared inside the successful upload condition.

Retrieval practice

Recall check 2 of 2

Data Dora says: answer from memory, then check your reasoning.

Q3Place each edge-processing responsibility where it lives so you can trigger a local action without confusing raw arrival, transformed evidence, and a decision.

ATime-Bound Stream Ingestion

BUnit-Normalized Local Transform

CContextual Pattern Detector

DAuditable Action Trigger

Show answer

Answer: A Place each edge-processing responsibility where it lives so you can trigger a local action without confusing raw arrival, transformed evidence, and a decision.

Print reference

Answers

Answer key.

A · The example returns minimum, maximum, average, deviation, count, and window end.
D · The buffer is cleared inside the successful upload condition.
A · Place each edge-processing responsibility where it lives so you can trigger a local action without confusing raw arrival, transformed evidence, and a decision.