ODSC West 2021: Data Science Performance Isn't What You Think

My talk at ODSC West 2021 was a deep dive of Data Science performance, utilizing instrumentation tools and benchmarks to reveal how Data Science performance is impacted by processor, system, and platform choices.

#DataScience #MachineLearning #AI #DataScienceWorkstation

00:00 Intro
00:32 Current State of Data Science Hardware
01:50 Where is the most time spent in Data Science?
02:59 Under the performance hood of Data Science Workloads
03:48 Why did my Laptop beat out a large Cloud instance?
04:30 VM "Split" issues
05:54 Core Count, Vectorization, and NUMA
08:11 Benchmark tests
10:44 NUMA and Network NUMA
12:14 Performance Hierarchy
13:15 Software Optimization and Performance Behavior
17:14 Memory Solutions
18:37 Building a Data Science system for the masses
20:59 Summary