Snowflakes at the Edge
| Autor: | Hassan H. Halawa, Matei Ripeanu, Hazem A. Abdelhafez, Karthik Pattabiraman |
|---|---|
| Rok vydání: | 2021 |
| Předmět: |
Edge device
Computer science 020206 networking & telecommunications 02 engineering and technology Stability (probability) Perspective (geometry) Computer engineering 020204 information systems 0202 electrical engineering electronic engineering information engineering Statistical analysis Enhanced Data Rates for GSM Evolution Snowflake Focus (optics) Edge computing |
| Zdroj: | EdgeSys@EuroSys |
| Popis: | While applications deployed at the edge often rely on performance stability (or, at a minimum, on a predictable level of performance), variability at the edge remains a real problem [4]. This study uncovers a surprising source of variability: intrinsic variability (in performance and power consumption) among edge platforms that are nominally identical. We focus on a popular platform designed for edge applications, the NVIDIA Jetson AGX, and aim to answer the following high-level questions through rigorous statistical analysis: (i) are the edge devices in our study statistically different from each other in terms of applications' runtime performance and power draw (although they are sold under the same product model and family)?, (ii) if the differences between these edge devices are statistically significant, what is the magnitude of these differences?, and (iii) do these differences matter from the application's perspective? |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|