Benchmarks - DSE-capstone-sharknado/main GitHub Wiki

BPR Speed Tests

Goal: Test BPR code to see where it scales the best. The results are intended to inform our decision making process on what technology to use to scale our analysis.

Code Tested (Local): https://github.com/DSE-capstone-sharknado/bpr/blob/master/sampling_test.py

Commit: a7f0dbc

Test Date: 2017-02-20

Spark Notebook: https://dbc-f6057a15-2f8d.cloud.databricks.com/#notebook/102637

Results:

13 Min - Alex's MBP

Julus' Macbook Pro: real 15m54.309s user 15m45.995s sys 0m4.847s

2 Node Spark Cluster: 1 loops, best of 3: 14min 33s per loop

Cluster Setup

On Demand Driver (15 GB Memory, 8 Cores, 1 DBU) x2

A Databricks Unit ("DBU") is a unit of Apache Spark processing capability per hour. Learn more Workers (30 GB Memory, 16 Cores, 2 DBU)

Related Slides