Benchmark

This benchmark is used to test the insert, load, scan of PostgreSQL and VidarDB (RocksDB style version).

We use TPC-H lineitem as the data source. Attention: The first attribute is implicit as the primary key for VidarDB engine. We also shuffle the data to avoid them coming in primary key order.

We have finished the testing with TPC-H 10G and 1G in Ubuntu 20.04 with PostgreSQL 12.3. You can follow the instruction in our repo of pgrocks-fdw to enable VidarDB storage engine.

Please don't hesitate to open an issue if you have a problem running with our benchmark.

Building

1. Install libpqxx which is the official C++ client API for PostgreSQL and 6.x.x version is preferred.

   • For Ubuntu:

     sudo apt-get install -y libpqxx-dev

   • For Others:

     You can follow the Building libpqxx to install.

2. Install VidarDB Engine static library which is needed for the benchmark tools.

3. Build all benchmark tools in the root directory of the repo:

   make all

Run benchmark

Before running benchmark, please ensure:

1. The PostgreSQL server has enabled VidarDB Engine.
2. The PostgreSQL server has no password and set trust in pg_hba.conf to 127.0.0.1/32 (currently the benchmark does not support password).
3. Best to clear the system cache (sudo sh -c "echo 3 > /proc/sys/vm/drop_caches") and restart the PostgreSQL server.

• Run benchmark scenario:

  # PLATFORM: benchmark database, optional values:
  #           pg: PostgreSQL
  #           fdw: VidarDB
  #           engine: VidarDB Engine
  #           tbam: lastest VidarDB
  #
  # SCENARIO: benchmark scenario, optional values:
  #           insert: line-by-line insertion
  #           load: for a bulk of lines
  #           scan: full scan
  #           getrand: get tuples by random primary keys  
  #           getlast: get tuples by last N primary keys 
  #
  # DATASIZE: benchmark data size, unit is GB.
  
  PLATFORM=fdw SCENARIO=insert DATASIZE=1 make run-benchmark