[FLINK-28751][Table SQL/Runtime] Optimize the performance of the json…

[Aitozi]

… functions

What is the purpose of the change

This PR is meant to improve the performance of the built in json functions. The default LRUCache used in the JsonPath CacheProvider heavily use the lock which bring the bad performance.

Brief change log

• Create a new JsonPathCache and set it to the CacheProvider when load the class

Verifying this change

The functionality is covered by the existing tests. I have not written a performance test for it, If needed, I will add one. I manually test the case with the production job, which will have 2~4 times performance improvement.

Does this pull request potentially affect one of the following parts:

• Dependencies (does it add or upgrade a dependency): (no)
• The public API, i.e., is any changed class annotated with @Public(Evolving): (no)
• The serializers: (no)
• The runtime per-record code paths (performance sensitive): (no)
• Anything that affects deployment or recovery: JobManager (and its components), Checkpointing, Kubernetes/Yarn, ZooKeeper: (no)
• The S3 file system connector: (no)

Documentation

• Does this pull request introduce a new feature? (no)

[flinkbot]

CI report:

• 625c09a Azure: SUCCESS

Bot commands

The @flinkbot bot supports the following commands:

• @flinkbot run azure re-run the last Azure build

[wuchong]

Do you have any benchmark between the two implementations?

[Aitozi]

I have not written a benchmark test for it now, I only run a production test which seen 2~4 times improved. Do you mean add an extra test for it in the flink-benchmark project? @wuchong

[Aitozi]

@wuchong I wrote a test with

• 4 threads
• 400 cache item

The result shows below:

Benchmark                 Mode  Cnt     Score     Error   Units
GuavaCacheBenchmark.get  thrpt   30  4480.563 ± 203.311  ops/ms
GuavaCacheBenchmark.put  thrpt   30  1774.769 ± 119.198  ops/ms

LRUCacheBenchmark.get  thrpt   30  441.239 ±  2.812  ops/ms
LRUCacheBenchmark.put  thrpt   30  350.549 ± 12.285  ops/ms

The results show that the new cache will bring 5~10 times throughputs

[Aitozi]

ping @wuchong

[Aitozi]

@wuchong any further comments ?

[wuchong]

LGTM.

[wuchong]

@flinkbot run azure

[Aitozi]

Rebase and force pushed