Optimize part size for checksummed read #315
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
monthonk
had a problem deploying
to
PR integration tests
GitHub Actions
Failure
monthonk
had a problem deploying
to
PR integration tests
GitHub Actions
Failure
monthonk
had a problem deploying
to
PR integration tests
GitHub Actions
Failure
monthonk
had a problem deploying
to
PR integration tests
GitHub Actions
Failure
monthonk
temporarily deployed
to
PR integration tests
GitHub Actions
Inactive
monthonk
had a problem deploying
to
PR integration tests
GitHub Actions
Failure
monthonk
had a problem deploying
to
PR integration tests
GitHub Actions
Failure
monthonk
had a problem deploying
to
PR integration tests
GitHub Actions
Failure
The prefetcher stores data received from each input stream as a part in the part queue structure. Usually, the part size is pretty big (8 MB or more) and the checksum validation always has to be done against an entire part even if we only read a small portion of that part. This makes checksummed read much slower than non-checksummed read. We could make it more efficient by making the part smaller or ideally align the part size to the read size so that we don't have to compute the checksum on unnecessary bytes. Signed-off-by: Monthon Klongklaew <monthonk@amazon.com>
monthonk
had a problem deploying
to
PR integration tests
GitHub Actions
Error
monthonk
temporarily deployed
to
PR integration tests
GitHub Actions
Inactive
monthonk
temporarily deployed
to
PR integration tests
GitHub Actions
Inactive
monthonk
temporarily deployed
to
PR integration tests
GitHub Actions
Inactive
monthonk
temporarily deployed
to
PR integration tests
GitHub Actions
Inactive
monthonk
temporarily deployed
to
PR integration tests
GitHub Actions
Inactive
monthonk
temporarily deployed
to
PR integration tests
GitHub Actions
Inactive
monthonk
temporarily deployed
to
PR integration tests
GitHub Actions
Inactive
mountpoint-s3/src/prefetch.rs
Outdated
Comment on lines
167
to
168
| let min_part_size = 128 * 1024; | ||
| self.preferred_part_size = length.max(min_part_size); |
There was a problem hiding this comment.
I think if you initialize preferred_part_size to 128KiB, this can just be
Suggested change
| let min_part_size = 128 * 1024; | |
| self.preferred_part_size = length.max(min_part_size); | |
| self.preferred_part_size = self.preferred_part_size.max(length); |
Also worth a comment on why we chose 128KiB.
There was a problem hiding this comment.
Also, should we set a maximum here, like 1MiB?
There was a problem hiding this comment.
For choosing 128k, it's the linux readahead size and seem to be a reasonable minimum value. I will update the comment, also happy to change if you have better suggestion.
I'm not sure we should put a maximum value though. If the read size is really big then we will have to combine data from multiple parts and the extend operation is quite expensive.
There was a problem hiding this comment.
but if we change the logic the use max value between last preferred_part_size and current length I think there should be a maximum, otherwise it will keep growing bigger.
There was a problem hiding this comment.
The effective maximum is the client's part size anyway, so probably we should enforce it here just to be explicit.
Signed-off-by: Monthon Klongklaew <monthonk@amazon.com>
monthonk
had a problem deploying
to
PR integration tests
GitHub Actions
Failure
monthonk
temporarily deployed
to
PR integration tests
GitHub Actions
Inactive
monthonk
temporarily deployed
to
PR integration tests
GitHub Actions
Inactive
monthonk
temporarily deployed
to
PR integration tests
GitHub Actions
Inactive
jamesbornholt
approved these changes
Jun 29, 2023
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.
The prefetcher stores data received from each input stream as a part in the part queue structure. Usually, the part size is pretty big (8 MB or more) and the checksum validation always has to be done against an entire part even if we only read a small portion of that part.
This makes checksummed read much slower than non-checksummed read. We could make it more efficient by making the part smaller or ideally align the part size to the read size so that we don't have to compute the checksum on unnecessary bytes.
By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license and I agree to the terms of the Developer Certificate of Origin (DCO).