[Java] Add performance benchmarks from SQL workloads

[asfimport]

To improve the performance of Arrow implementations. Some performance benchmarks must be setup first. 

In this issue, we want to provide some performance benchmarks extracted from our SQL engine, which is going to be made open source soon. The workloads are obtained by running an open SQL benchmarks TPC-H. 

Reporter: Liya Fan / @liyafan82
Assignee: Liya Fan / @liyafan82

Related issues:

• [Java] Provide light-weight arrow APIs (blocks)
• [Java] Add microbenchmark code to Java (is a child of)

PRs and other links:

• GitHub Pull Request apache/arrow#4198

_{Note: This issue was originally created as ARROW-5209. Please see the migration documentation for further details.}

[asfimport]

Liya Fan / @liyafan82:
I have sent a PR with the initial benchmark. The benchmark is extracted from running a query in TPC-H. The evaluation output looks like this:

 

Statistics for benchmark TPC-H Q1#Project & Filter:
Num of repeats = 10
Max duration = 0.64111ms
Min duration = 0.381478ms
Avg duration = 0.48580240000000013ms

 

If it is OK, I will provide more benchmarks concerning more TPC-H queries.