Published May 10, 2023 | Version v2
Software Open

SLASH - Serverless Apache Spark Hub

  • 1. Zurich University of Applied Sciences

Description

This folder contains SLASH, the research prototype implementation of
Serverless Apache Spark Hub.

To make it run, install the dependencies for OpenStack access, e.g.:
% sudo apt-get install python3-novaclient
Then, proceed with the script run.sh in the deploy folder.
Finally, use 'import slash' in your Spark Python scripts or Jupyter
notebooks.

SLASH consists of three components: The Python module 'slash' which
augments 'pyspark' magically with reactive autoscaling capabilities, the
'slashhub' which collects on-demand scaling actions to calculate
reactive autoscaling in conjunction with predictive ones, and the
'slashjobserver' which executes the calendar scheduling and forecasting.

A calendar file is expected as 'calendar' with dates or date ranges per
line. Furthermore, the functionality is determined by environment
variables: OS_* (see openstackclient.py for a list) and SLASH pointing
'slash' to the node running 'slashhub' on port 11111.

Files

Files (8.8 kB)

Name Size Download all
md5:c4445d5884c7c945c1b400126809adcf
8.8 kB Download