dice Benchmark Executor performance

The DSQL executor parses the DSQLQuery object and returns a result set of keys and/or values.

We currently have no benchmark of the performance characteristics of this operation.

Requirements

Create a benchmark with large number of keys in the store and validate query performance. It may be a good idea to check performance difference at ~1k, ~10k, ~100k keys in the store.
benchmark time taken and memory allocs using benchmem.
test cases should validate different kinds of queries (with/without ordering specified, with/without WHERE clause, with/without complex wildcard matchers)

Jul 31 '24 04:07 JyotinderSingh

HI @JyotinderSingh I can pick this up. Since I'm anyway going to dive deeper into the codebase, I can work on the test cases and benchmarking.

Jul 31 '24 04:07 AshwinKul28

HI @JyotinderSingh I can pick this up. Since I'm anyway going to dive deeper into the codebase, I can work on the test cases and benchmarking.

Assigned

Jul 31 '24 04:07 JyotinderSingh

HI @JyotinderSingh the plan I'm thinking of is to use a go-dice client and populate keys using the client.JSONSet() function and then write the benchmarks using the different scenarios suggested above. Does this sound okay?

Jul 31 '24 20:07 AshwinKul28

HI @JyotinderSingh the plan I'm thinking of is to use a go-dice client and populate keys using the client.JSONSet() function and then write the benchmarks using the different scenarios suggested above. Does this sound okay?

JSON support is not yet available on master

Jul 31 '24 21:07 JyotinderSingh

I have covered all the cases available in the executer_test.go to run the benchmark cases with the keys 100, 1000, 10000, 100000. Here's a report.

executerbenchmark_Nx.txt

PR: https://github.com/DiceDB/dice/pull/211

@arpitbbhayani @JyotinderSingh ^^

Aug 02 '24 19:08 AshwinKul28