How to use the results to get or calculate throughput and latency as the paper

[yolo-612]

i run ethereum's macro benchmark kv_store and run./run_bench.sh 2 4 2 10 and i got some results as follow, please help me how to use these data to calculate the throughput and latency as the paper?
Results: client_192.168.43.246: https://pastebin.com/uBvX84V3
client_192.168.43.162: https://pastebin.com/USRHRKUy
eth_log: https://pastebin.com/pZcUCTvp

how to get ?? thank you

[jonghwaC]

Did you solve it?

[CryptonQQQ]

I have the same problem. Can you give me the solution to the problem？

Did you solve it?

[yolo-612]

@jonghwaC no ，no one told me how to solve during this time.
Is your result similar to my result? for eth_log and client_result?

[jonghwaC]

I have a similar result.
But I want is throughput and latency.

So I am in small bank code analysis.
On line 175 in smallbank.cc, there is a portion that calculates throughput and latency.

"threads.emplate_back (StatusThread, sb, props [" dbname "], props [" endpoint "] ..."
Commenting out the above code shows Throughput but not Latency.

[CryptonQQQ]

I have a similar result.
But I want is throughput and latency.

So I am in small bank code analysis.
On line 175 in smallbank.cc, there is a portion that calculates throughput and latency.

"threads.emplate_back (StatusThread, sb, props [" dbname "], props [" endpoint "] ..."
Commenting out the above code shows Throughput but not Latency.

I am doing the same experiment as you, but there is no result, can we communicate?

[yolo-612]

@jonghwaC @CryptonQQQ I got some of result of latency has been 0.
you can get the throughput and latency of ycsb?

[jonghwaC]

@keruisiya @CryptonQQQ
I also have a latency value of 0 in smallbank
atomic latency (0); is defined and the variable is not used elsewhere.

I think I should use the latency defined in StatusThread.

My plan is to terminate StatusThread when all ClientThreads are terminated.
StatusThread will be modified to increase the globally defined latency value.

[yolo-612]

@jonghwaC ok！thanks! If you get the right results, please let me know！
For this benchmark, I also have some doubts，hope can communicate with you，
you can contact me at [email protected]

[haochenpan]

Hi, @keruisiya , @jonghwaC

Our team found this

https://github.com/ooibc88/blockbench/blob/f4563c49def905d6fadba9c031a6baf2187db6fe/src/macro/kvstore/aggregate.py.

Have you checked that and did that help?

Thanks,
Haochen

[listenfound]

This document is not used to analyze the results. It's a document in ycsb. The author modified ycsb, so it can't be used to analyze the results.

[haochenpan]

Hi @listenfound ,

You are right. see #49 .

After all, I don't think their way of calculating latency makes sense. See the discussion in #64. There could be a finer way of doing this.

Thank you for your discussion.

Best wishes,
Roger