Ganesh Vernekar
Ganesh Vernekar
So if we are able to use a singly linked list, we save 16bytes per series over the slice. Makes sense. Thanks for correcting my math! > Also I envisaged...
> FYI I did that in a different branch Can you update this branch to have a singly linked list for only the in-order part? It will be easier to...
> Sorry, it was for m-mapped chunks, not OoO, didn't touch that part at all. Oh, okay. Yeah let's skip the mmap chunks for now. I did not look at...
/prombench main I want to see how this latest change performs, especially with the conversion to slices at places.
There is some increased allocations and constant increase in CPU (the latter likely because of regular mmap checksm, and the increased allocations maybe?). Memory usage is not much affected. Will...
Oh, the query latency also seems to have increased a bit, which is not ideal. I feel we need to profile and investigate it a bit and improve the perf.
/prombench cancel
I plan to check this next week. Apologies for the delay. On the top of my head, I don't know what might be the issue. @rayrapetyan do the gaps go...
Also, @rayrapetyan do you see the gaps for all the metrics or only a subset of metrics?
Does doing the label cleanup step before calling `GetRef` work? Or does the extention of interface have other benefits as well? (I am not considering any modifications in labels.Labels right...