Dell PowerEdge R940xa GPU Database Acceleration on - Page 3

Table of contents Revisions...2 Acknowledgements ...2 Executive summary...4 1 Evolution of databases ...5 2 What is GPU acceleration and how does it apply to databases 6 2.1 Why GPU?...6 2.2 What database operations can run on GPU?...7 2.3 How do GPUs accelerate analytic workloads 7 2.4 What are some examples?...7 2.5 How can GPU database acceleration help other workloads like Machine Learning and Deep Learning? ........7 2.5.1 How does Brytlyt help with Machine Learning 8 2.6 Why keep CPU: GPU ratio of 1:1 in R940xa 10 2.6.1 Disk-IO bottleneck ...11 2.6.2 PCIe bottleneck ...11 3 The Dell EMC PowerEdge R940xa server ...12 3.1 CPU: GPU ratio ...12 4 The challenge with GPU Databases ...13 4.1 Overview of Brytlyt...13 4.2 Background on TPC-H Benchmarking [1]...14 4.3 Importance of TPC-H for Ad-Hoc Business Decision Making Activities 14 5 Brytlyt + Dell EMC PowerEdge R940xa Benchmarking 16 5.1 TPC-H benchmarking on PowerEdge R940xa with Brytlyt 16 5.2 NY TAXI Benchmarking on PowerEdge R940xa with Brytlyt 16 5.3 PowerEdge R940xa NY Taxi data runtimes...16 5.4 Previous Benchmark ...16 5.5 Reducing Time-to-Value for Data Analysts ...16 5.6 Use Cases for Brytlyt's GPU Database ...17 6 References ...18 3 GPU Database Acceleration on PowerEdge R940xa