Presentation: Architectures for Scalable Analytic Dashboards in Python with Dask Distributed and Dash | Dask Distributed Summit 2021

Time Zone: UTC

20 May 16:00 – 20 May 16:30 in Talks

Architectures for Scalable Analytic Dashboards in Python with Dask Distributed and Dash

Jon Mease

Audience level:: Intermediate

Description

Dash is a framework for developing analytic web apps in Python. This talk will describe Dash’s design, and how it enables efficient scaling to support large numbers of simultaneous users. Then, several architectures will be presented that can be used to combine the strengths of Dash with the strengths of Dask Distributed to create apps that scale to support large datasets and many users.

Abstract

Dash Overview

Project goals
Declarative and reactive development model
Client-side state model
Horizontal scalability

Dask Distributed Highlights

Distributed and local cluster configurations
Cluster persistence
Dataset publication
RAPIDS integration

Architecture 1:

Medium-size dataset
Local CPU parallelization
Horizontal scalability to support many simultaneous users

Architecture 2:

Medium-size dataset
Local GPU computation using RAPIDS
Horizontal scalability to support many simultaneous users

Architecture 3:

Horizontal scalability to support large dataset sizes
Distributed CPU or GPU computation
Horizontal scalability to support many simultaneous users

Future architecture:

Elastic scalability to accommodate variable dataset size and variable user traffic