Edge Computing vs Cloud for Low-Latency IoT Systems
Classified in Technology
Written on in 
English with a size of 3.63 KB
Issues with the Cloud
- Extremely low latency requirements: The whole process may need to take less than 10 ms. ossit l senser w neb3at l data 3al server w yeje jaweb w n3ml action (ex: icharet l ser) — (transliteration preserved; English: the sensor triggers, we send the data to the server, it replies and we perform an action, e.g., turn on the fan).
- Continuously sending unimportant data: e.g., a smart home sending temperature continuously.
- Offline decision and action requirements: Devices must take decisions and actions even if there is no Internet (e.g., oil pressure monitor).
Common cloud-related problems
- Data Transfer Latency: Some environments require near real-time decision making (e.g., medical systems, traffic monitoring, connected vehicles) and cannot accept back-and-forth transfer delays.
- Network Congestion: Large volumes of data will circulate over public networks (Internet), much of which does not need to be sent to the cloud (e.g., home temperature, light bulb status).
- Processing Bottlenecks: Most processing logic is implemented on cloud-based IoT platforms, even for simpler tasks such as data pre-processing.
- Centralized Decision Making: Processing and decision making is centralized in the cloud, which does not cope well with situations where decisions must be taken locally and quickly.
Solution: Edge Computing
Edge Computing offloads network and cloud infrastructures by moving processing and storage closer to where data is generated and used.
An option is to move part of the processing and data storage closer to where they are needed/generated/used, leveraging routers' and gateways' capacity to delegate tasks from the cloud.
Typical edge tasks
- Pre-processing of data
- Local decision making
- Calculations and local storage
- Real-time processing
Edge Computing extends cloud computing services to the edge of the network.
Cisco: In Fog Computing, devices communicate peer-to-peer to efficiently share and store data and take local decisions.
Cloud vs Fog (Edge) comparison
| Characteristic | Cloud | Fog / Edge |
|---|---|---|
| Latency | High | Low |
| Location of nodes | Internet | Edge of the network |
| Client / Server distance | Multiple intermediates | Direct |
| Location awareness | No | Yes |
| Security | Difficult — multiple vulnerabilities | Easier locally but risk of man-in-the-middle attacks |
Edge (Fog) Node Deployment
Edge (Fog) nodes need to deploy and run code. Two main approaches are possible:
Micro-services
- Require a container engine on the edge node.
- Deploy custom code (accessible as a service) into containers.
- Lightweight; deployment can be automated.
- Limited functionality: complex code is too expensive to run; limited to small-to-medium services.
IoT (edge) platforms
- Require deploying a platform on the edge node.
- Implement code using the platform; code gets automatically deployed.
- Heavy: can only run on edge nodes with high capacity (e.g., >= 2 GB RAM).
- Provides an advanced feature set: can build powerful services.