bench: Add benchmark test functionality for single node Tarantool topology #656

Kirill-Churkin · 2021-10-29T13:25:15Z

User setup Tarantool single node or cluster and try to understand «How many specific traffic Tarantool can handle on this hardware»
The same official things are for redis, postgresql and aerospike.

Cartridge bench module makes some load for Tarantool.

user@cartridge-cli % ./cartridge bench              
Tarantool 2.8.2 (Binary) f4897ffe-98dd-40fc-a6f2-21ca8bb52fe7  

Parameters:
        URL: 127.0.0.1:3301
        user: guest
        connections: 10
        simultaneous requests: 10
        duration: 10 seconds
        key size: 10 bytes
        data size: 20 bytes
Data schema
|       key             |       value
------------------------------------------
|       random(10)      |       random(20)
Benchmark start
...
Benchmark stop 

Results:
        Success operations:  1169481
        Failed  operations:  0
        Request count:  1170485
        Time (seconds):  10.000551801
        Requests per second:  117042

What has been done? Why? What problem is being solved?

I didn't forget about

Tests
Changelog
Documentation

Closes #???

cli/bench/bench.go

cli/bench/insert.go

cli/bench/config.go

cli/bench/bench.go

cli/bench/delete.go

cli/bench/bench.go

cli/bench/config.go

cli/bench/connect.go

cli/bench/delete.go

cli/bench/select.go

cli/bench/types.go

cli/bench/update.go

cli/bench/utils.go

cli/commands/bench.go

test/integration/bench/test_bench.py

cli/bench/config.go

LeonidVas · 2021-11-22T22:06:15Z

Hi!
You continue to ignore comments from #656 (review) :

I know that we didn't add a description to the functions before and we didn't write comments very often, but now we are trying to do it (see new projects like https://github.com/tarantool/tvisor , https://github.com/tarantool/tt), so I going to ask you to add function descriptions and comments to the code.
Decorate the commit message according to https://github.com/tarantool/tarantool/wiki/Code-review-procedure#commit-message (add a description to the commit message in the same way as in PR) and mark the commit as "Part of RFC: Cartridge bench tool for Tarantool #645"
Please, don't just resolve the conversation - write the answers, what desicion was received (and why, if necessary). This information can be used not only by the authors of the question, but also by other reviwers.
The length of a line of code must fit in 100 symbols. It provides easy to work with the code then sreen is split to 2 columns in the IDE.

Why are you removing operationsMap? The idea was fine (IMHO), just leave only insert for the first time.

cli/bench/bench.go

LeonidVas

Hi!
From #656 (comment) :

Decorate the commit message according to https://github.com/tarantool/tarantool/wiki/Code-review-procedure#commit-message (add a description to the commit message in the same way as in PR) and mark the commit as "Part of RFC: Cartridge bench tool for Tarantool #645"

I leave several minors comments, so you can accept or dispute them and start getting a second LGTM. I think it's logical to finish with Alexey (Up to you).

CHANGELOG.md

cli/bench/bench.go

cli/context/context.go

cli/commands/bench.go

LeonidVas · 2021-12-02T08:48:50Z

And yet one question: "Do you have test results on various hardware, or the test works in theory but has not been tested in practice?"

cli/bench/space.go

LeonidVas

rebase on master.
From commit message:
"User setup Tarantool single node or cluster and try to understand «How many specific traffic Tarantool can handle on this hardware»"
"Commit message should fit 72 symbols line width."
From bench: Add benchmark test functionality for single node Tarantool topology #656 (comment) : mark the commit as :Part of #645
It would be nice to get an answer to the question bench: Add benchmark test functionality for single node Tarantool topology #656 (comment)

mRrvz

Now the PR looks beautiful!

Please, fix @LeonidVas comments and CI.

filonenko-mikhail · 2021-12-06T11:18:43Z

And yet one question: "Do you have test results on various hardware, or the test works in theory but has not been tested in practice?"

macbook pro i7 (2.9Ghz) ddr3 at monday morning

Tarantool 2.8.2 (Binary) bf5c2dfd-93a1-4fab-9a34-e280d1c7de80

Parameters:
	URL: 127.0.0.1:3301
	user: guest
	connections: 10
	simultaneous requests: 10
	duration: 10 seconds
	key size: 10 bytes
	data size: 20 bytes
Data schema
| key                            | value
| ------------------------------ | ------------------------------
| random(10)                     | random(20)
Benchmark start
...
Benchmark stop

Results:
	Success operations: 1346103
	Failed  operations: 0
	Request count: 1348694
	Time (seconds): 10.000301
	Requests per second: 134865

User setup Tarantool single node or cluster and try to understand "How many specific traffic Tarantool can handle on this hardware" The same official things are for redis, postgresql and aerospike. Cartridge bench module makes some load for Tarantool. user@cartridge-cli % ./cartridge bench Tarantool 2.8.2 (Binary) f4897ffe-98dd-40fc-a6f2-21ca8bb52fe7 Parameters: URL: 127.0.0.1:3301 user: guest connections: 10 simultaneous requests: 10 duration: 10 seconds key size: 10 bytes data size: 20 bytes Data schema | key | value ------------------------------------------ | random(10) | random(20) Benchmark start ... Benchmark stop Results: Success operations: 1169481 Failed operations: 0 Request count: 1170485 Time (seconds): 10.000551801 Requests per second: 117042 Part of tarantool#645

Totktonada requested a review from LeonidVas October 29, 2021 15:43