Reduce mapped memory with many bulk clients #963
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
With this commit we share a parameter source for all bulk indexing clients per Rally worker. As all clients run in the same asyncio event loop they can also share a parameter source. This reduces the number of mmap system calls and thus virtual memory usage significantly: we only map the bulk data file(s) now only once per process instead of once per client. We can also better take advantage of prefetching as multiple clients within a process read now linearly from the mapped file. We have also changed the assignment rules slightly so successive client ids get assigned to a worker in order to read a continuous range of data.
danielmitterdorfer
added
enhancement
dliappis
approved these changes
Apr 27, 2020
| worker = host["workers"][worker_idx % len(workers)] | ||
| worker.append(client_idx) | ||
| host["worker"] = worker_idx + 1 | ||
| assert remaining_clients == 0 |
There was a problem hiding this comment.
Raise an exception or error out if this is not true?
There was a problem hiding this comment.
My intention was that this is a lightweight check for violation of an invariant due a logical error in the implementation (in the sense of "this thing should never fail"). Should that ever fail it would be easy to reproduce and that's why I used an assertion.
There was a problem hiding this comment.
Ok I see used as a debugging aid then ++
|
Thanks for the review! :) |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
With this commit we share a parameter source for all bulk indexing
clients per Rally worker. As all clients run in the same asyncio event
loop they can also share a parameter source. This reduces the number of
mmap system calls and thus virtual memory usage significantly: we only
map the bulk data file(s) now only once per process instead of once per
client. We can also better take advantage of prefetching as multiple
clients within a process read now linearly from the mapped file. We have
also changed the assignment rules slightly so successive client ids get
assigned to a worker in order to read a continuous range of data.