Skip to content

Latest commit

 

History

History
73 lines (46 loc) · 2.14 KB

README.md

File metadata and controls

73 lines (46 loc) · 2.14 KB
Raw

TVStore-benchmark

TVStore-benchmark is a benchmarking suite, with workloads and datasets, for fair comparisons between various time series store solutions, which might or might not support lossy compression. Currently, TVStore-benchmark supports TVStore, SummaryStore, Apache IoTDB and RRDTool. It constitutes partial contribution of our work on TVStore. For details please refer to our FAST'22 paper:

  • Yanzhe An, Yue Su, Yuqing Zhu, and Jianmin Wang. "TVStore: Automatically Bounding Time Series Storage via Time-Varying Compression." In 20th USENIX Conference on File and Storage Technologies (FAST 22). 2022. [pdf]

Prerequisites

To use TVStore-benchmark, you need to have:

  1. Java >= 1.8
  2. Maven >= 3.1

Installation

Run the following commands to install TVStore-benchmark.

git clone https://github.com/thulab/TVStore-benchmark.git
cd TVStore-benchmark
mvn clean install -Dmaven.test.skip=true

Usage

Take TVStore for example:

  1. Install, compile and run TVStore

    Please refer to TVStore.

  2. Modify configuration

    Modify conf/config.properties according to your needs. Note that parameter directory must be modified.

  3. Run TVStore-benchmark

    First, enter the binary directory.

    cd bin

    Then, run the corresponding command to meet your needs.

    • To measure throughput on synthetic data:

      ./ingest_synthetic_data.sh

    • To execute query process on synthetic data:

      ./query_synthetic_data.sh

    • To measure throughput on redd low frequency data:

      ./ingest_redd_low_data.sh

    • To execute query process on redd low frequency data:

      ./query_redd_low_data.sh

How to benchmark your solution

If you want to use TVStore-benchmark to benchmark your own time series store solution, you should complete a new class that extends cn.edu.tsinghua.tvstore.benchmark.store.Store and implement five necessary methods, including prepare, append, query, flush and finish.