Closed kpango closed 4 years ago
Oh, I'm sorry this problem does not happen now (and may did not happen) I can create index from scratch by only use ngtd HTTP API.
On the first phase I try to add search (or searchbyid) benchmark code.
For it I think it is useful to add dummy db which does not use any real db to ignore db specific performance differences or change. Do you have any concern about it?
As a precaution, I think that it is necessary to make the vector data as unbalanced as possible so as not to bias the vector data as a benchmark.
I plan to use JMeter for benchmark (and it is not run in CI), is it right? Or did you mean write BenchmarkXXX function in xxx_test.go?
I mean you should write BenchmarkXXX function code in xxx_benchmark_test.go. After that you should write e2e test code using JMeter or Vegeta or ApacheBench. I strongly recommend create e2e test package inside the ngtd using vegeta sdk.
Oh I see. I'm a Go newbie, so it may takes time.
It's okay, I'm proficient in Golang. I can review your code.
Thank you!! 😄
Although this is a digression, I'm already headed to the next version of ngtd (clusterize & native core api). If you have no time right now, you can cooperate after publishing the next version.
I'd like to try this job.
I think it is easy to use ngtd insert api for setup index, but ngtd can't start up without NGT index. So, at first we should fix this problem. I'll make this issue with detail info.