Chronix benchmark

A benchmark for Chronix and other time series databases (TSDB)

Requirements

• Server:

  • Ubuntu 12.04 or later. (any Linux distro with installed GNU Coreutils should work, but tested is only Ubuntu)
    Mac OS X El Capitan is also supported.
  • 16GB of RAM
  • Installed Java 8 or later
  • Installed and running docker
  • SSD is recommanded.

• Client:

  • Ubuntu 12.04 or later.
    Mac OS X El Capitan is also supported.
  • 8GB of RAM (recommanded 16GB)
  • Installed Java 8 or later

Installation

#####Server:

• Download the benchmark server part(link) to your server.
• Unzip the downloaded file to a folder of your liking.
• In terminal go to that folder and start the server:
  java -Xmx8g -jar server.jar server DropWizardServer.yaml

#####Client:

• Download the benchmark client part(link)
• Unzip the downloaded file to a folder of your liking.
• In terminal go to that folder and setup the client:

  • Add a TSDB to the benchmark:
    java -jar client.jar setup add [server] [tsdbName] [hostPort:containerPort] ["additionalDockerOptionsString"] [<TSDB>.jar] [dockerFilesFolder]

    example: (NOTE: paths must be absolute) java -jar client.jar setup add localhost yourTsdbName 4711:4711 "-v /dataFolderOfYourTSDBinContainer" /path/yourTsdb.jar /path/yourDockerFilesFolder java -jar client.jar setup upload localhost java -jar client.jar build localhost yourTsdbName

Integrating your own TSDB

• Create a Dockerfile which provides the environment for your TSDB and also give instructions to the Dockerfile to install and start the TSDB. (Examples can be found here)
• Add DatabaseInterface.jar(link) to a Project and implement the BenchmarkDataSource interface with a class for your TSDB. (example for chronix can be found here)
• Build a fat jar of your BenchmarkDataSource interface implementation, for example with shadowjar.
• Provide the necessary docker files and the fat jar of your TSDB to the client: see "Installation Client"

Preparing the import data

• Data has to be in csv format (could also be gzipped) and has to conform to:
  • File format: /measurement/host_process_group_metric.csv.(gz)

  •      file header line: Date;metricName1;metricName2;...
    

  •      file data:        2015-03-04T13:59:46.673Z;0.0;0.0;...  
    

• In terminal convert the csv data to json files:
  • java -jar client.jar convert [csvFilesDirectory1] [csvFilesDirectory2] ...

Importing data

• In terminal import the converted data to the TSDBs on the benchmark server: (adjust batchSize if you run out of memory)
  • java -jar client.jar import [server] [batchSize] [fromFile] -t [tsdbName1] -t [tsdbName2] ... -d [directoryToImport1] -d [directoryToImport2] ...
  • example:
    java -jar client.jar start localhost someTsdb java -jar client.jar import localhost 25 0 -t someTsdb -d /path/chronixBenchmark/timeseries_records/someFolder

Generating benchmark data

• From the previously imported data generate a set of random time series which will be queried in later benchmarks
  • java -jar client.jar generate [timeSeries_count_per_query] [total_count_of_queries] [timeSeries_count_per_rangeQuery]
  • example:
    java -jar client.jar generate 400 1600 50

Running the benchmark

• Start the TSDB(s) you want to benchmark: (example)
  • java -Xmx8g -jar client.jar start localhost someTsdb
• Start the benchmark: (example)
  • java -Xmx8g -jar client.jar benchmark localhost someTsdb

Benchmark analysis

• Analyze time series: (takes some time depending on your data)
  • java -jar client.jar stats timeseries
• Generate bar diagrams: (under /path/chronixBenchmark/statistics/bar_plots
  • java -jar client.jar stats plot