Site Navigation : Introduction of Home Edge Project | Home Edge Platform Architecture and Modules | Technical Requirements Details and Use Cases (TBD)
Definition : Technologies of Home Edge Project
- Technologies that compute and proceed user data in real-time by using distributed edge devices at home network.
- BENEFIT : Reinforcement of user privacy, low latency
- Performance examples when employing Home Edge Project technologies
| Smart TV | Home Edge ** | |
|---|---|---|
| * # of Cores | 4 | 28 |
| * Memory | 2 GB | 12 GB |
Service examples | Voice recognition for device control (Required anticipated size of ML Model : 20 MB) | Voice recognition for contents searching (e.g. media, retail, etc.) (Required anticipated size of ML model : 100 MB) |
* Given number is an example referring from the product specification available from the Market (TV, Refrigerator, Air Conditioner, Speaker, and Mobile Phone).
** It is assumed that there are 5 edge devices at home enabling the Home Edge Project technologies in this example.
Highlights : Required Technologies
- Edge Orchestration : for deploying / searching / and managing services for distributed edge devices at home.
| Key Features | Description |
|---|---|
| Edge Device / Service Discovery | Discovery on edge devices and their services (e.g. voice recognition, device control, etc.) |
| Resource Capability Exchange | Exchange the available (computing) resource information (e.g. CPU, GPU, NPU, Storage, type of connected devices, etc.) |
| Topology Decision | Decision on Master / Slave device roles, devices that are able to interoperate with cloud |
| Real-time Monitoring | Offers network connection status (e.g. power on / off), services based on (computing) resource at home |
- Distributed & Parallel Machine Learning : for enabling inference / model learning at home with low latency.
| Key Features | Description |
|---|---|
| Data Parallelism | Distribution and Parallel ML through data partition into multiple edge devices at home |
| Service Parallelism | Distribution and Parallel ML through multiple services into multiple edge devices at home |
| Model Parallelism | Distribution and Parallel ML through knowledge model partition into multiple edge devices at home |