Edge computing processes data near where it is produced — at or close to the device or sensor — instead of shipping everything to a distant data center, reducing latency and bandwidth use.1
Overview
In a purely centralized model, devices send raw data to the cloud and wait for a response. Edge computing moves some of that work outward: a small computer near the source filters, aggregates, or acts on data locally, sending only summaries or alerts upstream. This cuts round-trip latency, saves bandwidth, and keeps working when the network link is slow or down.
It is closely tied to the internet of things, where many sensors generate more data than is practical to forward whole. The edge node may be an industrial gateway, a cell-tower micro data center, or a hobbyist single-board computer — the defining trait is proximity to where data is born, not any particular hardware.
Trade-offs
Where computation happens is a balance, and edge and cloud each win on different axes:
| Property | Edge | Central cloud |
|---|---|---|
| Latency | Low (local) | Higher (round trip) |
| Bandwidth use | Low (summaries) | High (raw data) |
| Compute power | Modest | Effectively unlimited |
| Works offline | Yes | No |
| Storage & analytics | Limited | Deep |
The two are complementary, not rivals: keep the reflex-fast, high-volume work at the edge and let the cloud do the heavy, long-horizon lifting.
Where it fits
Edge computing complements the cloud rather than replacing it — heavy storage and analytics stay central while time-sensitive or high-volume processing happens locally. A content delivery network is edge computing for content. GopherTrunk is naturally an edge workload: a Raspberry Pi by the antenna decodes RF on the spot and forwards only the decoded calls, rather than streaming raw IQ across the network.
Sources
-
Edge computing — Wikipedia, on the edge computing model. ↩