Large Cluster Load Test

Overview:

• This is a Benchmark intended to test mirror datagrid under load

Project structure:

The project consist of the following pu's:

• mirror
• space (defaults topology is 2,1)
• crew-members-feeder - populating the space with 100 crew members with each deployment (evenly spread to all datagrids)
• flights-feeder - reading the crewmembers from space, then populating the space with 1000 with each deployment (evenly spread to all datagrids)
• benchmark - client that constantly queries the space for flights and crew members every 1 sec and print a summery of the queries duration every 30 sec

Running instructions:

1. Verify in pom.xml that correct gsVersion used
2. Build the project (cd into project directory and run: mvn clean install)
3. Deploy the pu's - here you have 2 options:

• option1(this option for local deployment only): run: mvn xap:deploy from the project root directory to deploy all the pu's in one line using mvn plugin

• option2: deploy each pu manually(gs-ui/webui/cli) at this order:

  • mirror
  • space
  • crew-members-feeder
  • flights-feeder
  • benchmark

  you don't have to undeploy the feeders because they are not active after they write the objects to space(1 time only)

4. Have a look at the gsc that the benchmark pu sitting at to see intermediate reports like this:

-------------Intermediate Summery Start-------------

Summery info:

Total num of queries: 2,310

min query time: 1 ms (1,087,377 ns)

max query time: 43 ms

avg query time: 1 ms (1,976,335 ns)

num of exceptional queries: 0

avg exceptional queries time: 0

Thrown exceptions by type:

-------------Intermediate Summery End-------------

this intermediate report is printed every 30s

5. To stop the benchmark and see final summary undeploy the benchmark pu and have look at his gcs