Artifact Overview

This repository contains the code for our VLDB'21 paper, "Shashank Gugnani, Arjun Kashyap, and Xiaoyi Lu. Understanding the Idiosyncrasies of Real Persistent Memory".

We provide the following in this repository:

Source for PMIdioBench, the micro-benchmark suite proposed in Section 3
Source for ring buffer designs proposed in Section 6
Source for linkedlist designs proposed in Section 6
A set of scripts to reproduce idiosyncrasy, numa study, and lock-free results presented in Sections 4, 5, and 6, respectively
A set of scripts and data files to plot all graphs in the paper

Directory Layout

Bench/ contains the source for idiosyncrasy benchmarks
ring-buffer/ contains the source for TX and TX-free ring buffer designs
linkedlist/ contains the source for TLog and Log-free linkedlist designs
Data/ contains Excel files containing data points for all experimental analysis in the paper
Graphs/ contains PDF files of all graphs in the paper
Scripts/ contains scripts to plot all graphs in the paper

System Requirements

Linux x86_64 server
PMEM formatted as DAX FS
At least 2 NUMA nodes (required for I4 and NUMA study experiments)

Steps to reproduce idiosyncrasy and NUMA study results

Follow instructions in Bench/README.md
Copy idiosyncrasy results to Data/pmem-idiosyncrasies.xlsx
Copy NUMA study results to Data/numa-study.xlsx

Steps to reproduce ring buffer results

Follow instructions in ring-buffer/README.md
Copy results to Data/lock-free-results.xlsx

Steps to reproduce TLog and Log-free linkedlist results

Follow instructions in linkedlist/README.md
Copy results to Data/lock-free-results.xlsx

Steps to reproduce SOFT and Link-free linkedlist results

git clone https://github.com/shashankgugnani/Efficient-Lock-Free-Durable-Sets.git
cd Efficient-Lock-Free-Durable-Sets
Scripts/testLists.sh
The script will run all experiments and print throughput results. Latency results will be written to adr-latency.log and eadr-latency.log.
Copy results to Data/lock-free-results.xlsx

Steps to plot all graphs in the paper

Prerequisites: Python installation with jupyter notebook, matplotlib, and xlrd packages
Open Scripts/graphs.ipynb in jupyter notebook
Run the complete script
Plotted graphs will be available in the Graphs directory

Reference

Please cite PMIdioBench in your publications if it helps your research:

@article{gugnani-vldb21,
  author = {Gugnani, Shashank and Kashyap, Arjun and Lu, Xiaoyi},
  title = {Understanding the Idiosyncrasies of Real Persistent Memory},
  journal = {Proceedings of the VLDB Endowment},
  volume = {14},
  number = {4},
  pages = {626--639},
  year = {2021}
}

padsys/PMIdioBench