Redis Vector Index Sizing Simulator

Summary

Tool for generating VSS scenarios and then analyzing memory usage

Features

• Calculates memory footprint of a Redis vector index scenario

Prerequisites

• Python
• Docker

Installation

1. Clone this repo.

git clone https://github.com/Redislabs-Solution-Architects/vss-sizer.git && cd vss-sizer

2. Install Python requirements (either in a virtual env or global)

pip install -r requirements.txt

3. Start Redis Enterprise Docker environment

./start.sh

Usage

Options

• --url. URL connect string. Default = redis://default:redis@localhost:12000
• --nkeys. Number of keys to be generated. Default = 100,000.
• --indextype. Vector Index Type. Default = flat.
• --metrictype. Vector Metric Type. Default = l2.
• --floattype. Vector Float Type. Default = f32.
• --vecdim. Vector Dimension. Default = 1536.
• --vecm. HNSW M Param. Default = 16.

Execution

python3 vss-sizer.py --nkeys 100000 --objecttype hash --indextype flat --metrictype cosine 
--floattype f32 --vecdim 1536

Output

Sample output for the test above.

Vector Index Test
 
*** Parameters ***
nkeys: 100000
objecttype: hash
indextype: flat
metrictype: cosine
floattype: float32
vecdim: 1536
 
*** Results ***
index ram used: 606.98 MB
data ram used: 812.24 MB
index to data ratio: 74.73%
document size: 7416 B
execution time: 4.37 sec