Arcon - GitHub Pages
Arcon
Continuous and Deep Data Stream Analytics
@Birte19
Max Meldrum, Klas Segeljakt, Lars Kroll Paris Carbone, Christian Schulte, Seif Haridi
Outline
? Project Introduction ? Arc IR and Compilation Pipeline ? Demo (Frontend, IR, CodeGen, Execution) ? Conclusions and Future Work
2
Motivation
?Many Frameworks/Frontends for different needs ?(ML Training & Serving, SQL, Streams, Tensors, Graphs)
Tensor
Feature
AI
RL
Relational Data
Streams
Dynamic Graphs
ML
Simulation
Reasoning
3
Silos in Data Pipelines
Tensorflow
R, Pandas, scikit-learn,
...
Flink / Beam / Trill
Timely Data flow
Network
Ray
data lake
Spark/Giraph
- Impedance Mismatch ( e.g., types, guarantees, state etc.)
- Excessive IO/ Data Movemenent of intermediate results
- Isolated HW Execution - No cross-framework optimisation
4
The Arcon System
Arc IR Compiler Arcon Runtime
5
................
................
In order to avoid copyright disputes, this page is only a partial summary.
To fulfill the demand for quickly locating and searching documents.
It is intelligent file search solution for home and business.