TraceMesh: Scalable and Streaming Sampling for Distributed Traces

This is the replication package for [CLOUD'24] TraceMesh: Scalable and Streaming Sampling for Distributed Traces.

In this paper, we propose TraceMesh, a scalable and streaming trace sampler.

Repository Organization

├── docs/
├── datasets/online_boutique
│   ├── train.csv # The training trace dataset
│   ├── test.tar.gz # The compressed test trace dataset
│   └── test_label.txt # The id of annotated uncommon traces
├── src/
│   ├── DenStream/ # The modified implementation of DenStream
│   ├── path_vector.py # The implementation of path vector construction
│   ├── sketch.py # The implementation of sketch construction
│   ├── read_trace.py # The implementation of trace reading
│   └── trace_mesh.py # The main interface of TraceMesh algorithm
├── requirements.txt
└── README.md

Quick Start

Installation

1. Install python >= 3.9.

2. Install the dependency needed by TraceMesh with the following command.

​pip install -r requirements.txt

Data Preparation

We have collected, processed and cleaned the trace data from the online_boutique system for demonstration purposes. The trace data is available in the form of CSV files within the datasets/ directory. By replacing the data in the same format, one can seamlessly adapt TraceMesh to another system.

To begin, please decompress the test.tar.gz file in order to obtain the original test trace file named test.csv.

cd datasets/online_boutique
tar -xzvf test.tar.gz

Demo Execution

• Run TraceMesh

cd src
python trace_mesh.py

• Explanations of parameters:

usage: trace_mesh.py [-h] [--sketch_length SKETCH_LENGTH] [--eps EPS] [--data_path DATA_PATH] [--dataset DATASET] [--budget BUDGET]

options:
  --sketch_length SKETCH_LENGTH     Length of the sketch (default: 100)
  --eps EPS                         Epsilon value for clustering (default: 0.1)
  --data_path DATA_PATH             Path to the dataset (default: "../datasets/")
  --dataset DATASET                 Name of the dataset (default: "online_boutique")
  --budget BUDGET                   Budget for the sampling (default: 0.01)