Developing comprehensive performance metrics and optimization strategies for edge computing infrastructures.
Edge faces issues with efficiency, compute availability, and understanding task assigned or requested.
Benchmarks are useful but, existing research is mainly focused on Edge-Cloud pipeline.
Create specific benchmarks based on compute metrics to aid scheduler in Edge Platforms.
Our comprehensive approach to edge computing performance analysis
Collect data on the compute metrics that are vital in running popular applications in Edge Platoforms.
Intelligent workload distribution and caching strategies for minimal latency.
Efficient allocation algorithms for distributed edge computing systems.