apinkin / search-perf Goto Github PK

Searcher performance comparison suite between Solr, ElasticSearch and Sensei

1. Setup

By default the search-perf will load and query the data based on the data/cars.json dataset

The following setup steps are required to prepare the Sensei DB for testing:
Clean all the indexes, copy configs/sensei/schema.xml into the Sensei config folder

The following setup steps are required to prepare the Solr for testing:
Clean all the indexes in the data folder, copy configs/solr/schema.xml into the Solr conf folder

Elastic search doesn't require any setup steps

2. Loading docs into the DB:
It does make sense to do the performance testing when the documents are kept adding to the db

For the Sensei just use com.senseidb.gateway.file.LinedFileDataProvider to load docs from the big file
For Solr execute  bin/loadIndex script. Please uncomment the section specific to Solr
For Elastic Search execute  bin/loadIndex script. Please uncomment the section specific to Elastic Search
The script launches the java program with the following command line params:
type   url documentFilePath numOfThreads
eg 
- elastic http://localhost:9200/cars/car data/cars3m.json 5
- solr http://localhost:8983/solr/update/json/?commit=true data/cars3m.json 5

numOfThreads - number of threads, that are used to post documents to the DB


3. Do the performance test

For Sensei :
bin/perfTest configs/test-sensei.properties
For Solr
bin/perfTest configs/test-solr.properties
For Elastic search
bin/perfTest configs/test-elastic.properties

configs/test-xxx.properties file contains the configuration needed to launch the performance test. The majority of the parameters are self-explaining. 
schemaPath and dataFilePath specify the sensei schema and the file containing documents in the json format. These are needed to generate dynamic search queries

The result of the test will be printed on the console. Essential logs are located in the ./report.log file