Compute Placement Calculator

Compare edge, fog, and cloud placement for an IoT workload and see the tradeoffs move.

animation

edge-computing

fog-computing

cloud

latency

architecture

Learner-ready compute placement workbench with scenario presets, animated edge-fog-cloud flow, latency and bandwidth estimates, scoring, recommendations, and mobile-ready learning support.

Edge Fog Cloud Latency model

Place the compute where the IoT decision can actually work.

Pick a workload and watch one event move from the device toward edge, fog, and cloud options. The calculator scores each placement using response time, privacy, bandwidth, autonomy, capacity, and cost.

Edge placement Best current fit

49 ms Estimated decision latency

11.9 GB/day Raw workload generated

0.6 GB/day Traffic sent toward cloud

Edge means on-device.

Use it when the device must act immediately, keep private data local, or survive a lost connection. The limit is local CPU, memory, battery, and update management.

Fog means near the devices.

A gateway or site server can aggregate many devices, run larger models, and keep WAN traffic smaller while still avoiding a long cloud round trip.

Cloud means centralized scale.

Cloud placement is strong for fleet-wide analytics, long history, training, and elastic compute. It is weaker for urgent control loops and raw sensitive streams.

Animated placement path

Step through how the same event behaves at each possible compute layer.

Stage 1 Capture event

Sensor reading or media frame is created.

Stage 2 Try edge

Can the device decide locally?

Stage 3 Try fog

Can a nearby gateway decide in time?

Stage 4 Try cloud

Can central compute meet the target?

Stage 5 Place compute

Choose the best scoring layer.

Device source

12 devices, 30 events/min each

6.0/s

Edge device

Immediate local processing with limited compute budget.

49 ms

Fog gateway

Nearby site compute after one local upload.

36 ms

Cloud service

Remote elastic compute after WAN delay.

77 ms

raw event

Stage 1 of 5: Capture event

The device creates a workload event. The important question is whether the response must happen before data crosses the access link.

Recommended: Edge placement

Edge currently balances local response, privacy, autonomy, and reduced cloud traffic for this workload.

Workload controls

Change the scenario or inputs and the placement updates immediately.

Devices 12 More devices increase event rate, storage, cost, and fleet coordination load.

Events per device per minute 30 This is the raw event rate before filtering or aggregation.

Payload per event

Decision latency target

Compute complexity 3/10 Higher values represent filtering, ML inference, computer vision, or heavier feature extraction.

Access bandwidth

Privacy sensitivity 5/10 Higher sensitivity favors local processing or aggressive redaction before cloud storage.

Connection availability 97% Edge can keep acting offline; fog can keep a site running; cloud needs the path to be up.

How the latency estimate works

The calculator uses a teaching model, not a vendor quote. The raw upload time is payload KB x 8 / Mbps milliseconds. Edge avoids the raw upload before the decision. Fog adds one local upload and gateway processing. Cloud adds the raw upload plus a WAN delay before cloud processing.

How the placement score works

The score blends latency, privacy, bandwidth, autonomy, compute capacity, and relative cost. A score near 100 means the placement fits the selected workload well. A low score means at least one operational constraint is likely to dominate.

Rules of thumb

Use edge for urgent local action, offline autonomy, and sensitive raw data.
Use fog when several devices need nearby shared compute, aggregation, or a site-level decision.
Use cloud for long-term history, fleet optimization, heavy elastic jobs, and cross-site learning.
Use hybrid designs when edge or fog makes the decision and cloud stores summaries or trains models later.