This repository contains suplementary data for our paper "B-Trees Are Back: Engineering Fast and Pageable Node Layouts". The B-Tree implementation contained is to help others reproduce our findings. It is not suitable for production use.
The main artifacts produced are benchmarking binaries produced from the makefile target ycsb-all.
Each binary implements benchmarking of one data structure configuration and follows the naming scheme <config>-<debug_assertions><optimization_level>-ycsb.
config specifies corresponds to one of the configurations in the named-configs directory.
There are configurations for the B-Tree with various optimizations enabled, as well as for different in-memory data structures.
Each binary is configured via a set of environment variables, which are processed in ycsb2.cpp
By default, all B-Trees use a node size of 4KiB.
The script build_var_page_size.sh is used to build B-Tree configurations with different node sizes.
Examples can be found as comments in the script file.
Each subdirectory of the R directory contains python scripts to generate benchmarks, results of said benchmarks as csv files, and R code to analyze the results.
The output of each python script is a sequence of program invocations, intended to be piped into GNU parallel.
The csv file containing the results is generally named after the python script that produced it.