OPERATIONALIZING MACHINE LEARNING USING GPU 1 ACCELERATED, - PowerPoint PPT Presentation

OPERATIONALIZING MACHINE LEARNING USING GPU 1 ACCELERATED, IN-DATABASE ANALYTICS

Why GPUs? Performance Increase A Tale of Numbers Infrastructure Cost Savings 100x 75% Performance Costs 100x gains over traditional 75% reduction in infrastructure costs, licensing, RDBMS / NoSQL / In-Mem staff, etc. Databases 3000 Cores More with Less vs Modern GPUs can consist of Increase performance, 32 up to 3000+ cores compared throughput, capability while to 32 in a CPU minimizing the costs to support the business 2

Why a GPU Database? • Leverage Innovations in CPUs and GPUs • Single Hardware Platform • Simplified Software Stack 3

What are AI, ML, and Deep Learning? AI Deep Learning ML Predict y using function on data x 4

AI/ML/Deep Learning Cheat Sheet No shortage of techniques and programing languages 5

ML Cheat Sheet Python and SQL cover almost all the algorithms in that scary spider and Kinetica supports all Python libraries! 6

ML/AI/Deep Learning Lifecycle 7

ML/AI/Deep Learning Lifecycle • Create, extract, transform, and process big data: batch and streams • Apply ML to data. • Model pre-processing • Model execution • Model post-processing • Within an ecosystem of general analytics • Supporting a range of human and machine consumers 9

Typical AI Process: High Latency, Rigid, Complex Tech Stack BUSINESS USERS DATA SCIENTISTS ??? ENTERPRISES SPECIALIZED AI/ DATA STRUGGLE TO SCIENCE TOOLS MAKE AI MODELS AVAILABLE TO BUSINESS EXTRACT SUBSET EXTRACTING DATA FOR AI IS EXPENSIVE AND SLOW 9

Kinetica: A More Ideal AI Process BUSINESS USERS Monte Carlo Risk Custom Function 2 Custom Function 3 DATA SCIENTISTS API EXPOSES CUSTOM FUNCTIONS WHICH CAN BE MADE AVAILABLE TO BUSINESS USERS UDFs 10

Current Inefficient Use of Python • Interpreted • Single threaded = • Clean, transform • Flow: for each member python • Pre-process • Model execute • Post-process 11

Optimized SQL and Python UDF with Kinetica • Pre-process • Binary executable code SQL • Superior optimization • declarative SQL = • Model execute UDF • Only essential imperative model code python • Not relational set processing • Post-process • Binary executable code SQL • Superior optimization • Declarative SQL 12

Comprehensive Solution Architecture Major U.S Retailer Fast Streaming Fast Analytics Apache Tomcat Applications Servers Projects Projects Massive Stream • Spring Endpoint oriented architecture Massive Fast Ingestion • Analytics Horizontal elastic scaling KINETICA: 10 Node Cluster Worker Worker Head 1 9 Node Full Model Pipeline 1 Various Various Prompts ETL/ELT ETL/ELT Project Full Model Pipeline N Fact and dimensions tables for various Use Cases Billions of rows 13

Use Case Example

MNIST: Simple Image Processing Use Case A Parametric ModelPython Using TensorFlow Model Training • Set of image files stored in Kinetica Database Table • Python UDF in Kinetica using TensorFlow Model Serving • Python UDF in Kinetica using TensorFlow • Input = table TFModel table. • Output = table mnist_inference_out Model Analytics • SQL! 15

Model Training & Inference Da Data Model: MPP Shar arding Machine 0 Machine 0 Rank 0 Rank 0 Tom 3 Tom 1 Tom 7 Tom 2 Tom 5 Tom 6 Tom 0 Tom 4 Table Table Table Table Table Table Table Table mnist_training mnist_training UDF mnist_training mnist_training mnist_training mnist_training mnist_training mnist_training Shard 3 Shard 1 Shard 7 train_nd_udf.py Shard 2 Shard 5 Shard 6 Shard 0 Shard 4 Table Table Table Table Table Table Table Table TFModel TFModel TFModel TFModel TFModel TFModel TFModel TFModel Shard 3 Shard 7 Shard 1 Shard 2 Shard 5 Shard 6 Shard 0 Shard 4 Table Table Table Table Table Table Table Table mnist_inference mnist_inference mnist_inference mnist_inference mnist_inference mnist_inference mnist_inference mnist_inference Shard 3 Shard 7 Shard 1 Shard 2 Shard 5 Shard 0 Shard 6 Shard 4 UDF UDF UDF UDF UDF UDF UDF UDF Table Table Table Table Table Table Table Table mnist_inference_out mnist_inference_out mnist_inference_out mnist_inference_out mnist_inference_out mnist_inference_out mnist_inference_out mnist_inference_out Shard 3 Shard 1 Shard 7 Shard 2 Shard 5 Shard 6 Shard 0 Shard 4

Thank You! Come get your copy of the O’Reilly Book at Booth G.01! info@kinetica.com