Skip to content

[HotStorage'24 Best Paper] Can Modern LLMs Tune and Configure LSM-based Key-Value Stores?

Notifications You must be signed in to change notification settings

asu-idi/ELMo-Tune

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

4 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

ELMo-Tune ([HotStorage'24 Best Paper] Can Modern LLMs Tune and Configure LSM-based Key-Value Stores?)

🏆HotStorage'24 Best Paper - Can Modern LLMs Tune and Configure LSM-based Key-Value Stores?
Paper URL: https://doi.org/10.1145/3655038.3665954

Features

This project will run a series of tests using the db_bench tool. The tests will be run using the default configuration and a series of configurations that will be determined by the research. The results of the tests will be compared to determine the best configuration for RocksDB when using ELMo-Tune.

Prerequisites

This project requires Python 3.6 or higher. The following dependencies are required:

# Instructions for Ubuntu 20.04
# Install dependencies
apt-get update && apt-get install -y build-essential libgflags-dev libsnappy-dev zlib1g-dev libbz2-dev liblz4-dev libzstd-dev git python3 python3-pip wget fio 

# Install and Build RocksDB 8.8.1
wget https://github.com/facebook/rocksdb/archive/refs/tags/v8.8.1.tar.gz
tar -xzf v8.8.1.tar.gz
cd rocksdb-8.8.1
make -j static_lib db_bench

git clone https://github.com/asu-idi/ELMo-Tune
cd ELMo-Tune

# Install requirements
pip install -r requirements.txt

Setup

To run the tests sucessfully, some variables need to be defined.

# You need OpenAI's API to run the code sucessfully. 
export OPENAI_API_KEY=<api key>

Additionally, set the DB_BENCH_PATH in utils/constants.py along with any other paths required for your system setup.

How to use

To run the tests, run the following command:

# e.g. Run a random write (fillrandom) test with the db stored in the '/data/gpt_project/db' folder and with output in the './output' directory 
python3 main.py --workload=fillrandom --device=data --output=./output --num_entries=10000

# You can explore the options using the --help command (or using the constants.py file)
#  -c --case            CASE            Specify the case number
#  -d --device          DEVICE          Specify the device
#  -t --workload        WORKLOAD        Specify the test name
#  -v --version         VERSION         Specify the version of RocksDB
#  -o --output          OUTPUT          Specify the output path
#  -n --num_entries     NUM_ENTRIES     Specify the number of entries
#  -s --side_checker    SIDE_CHECKER    Specify if side checker is enabled

You can alternatively also use the Docker environment that can be built using the Dockerfile in the docker folder.

About

[HotStorage'24 Best Paper] Can Modern LLMs Tune and Configure LSM-based Key-Value Stores?

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published