Vortex Layouts - Drivers #1914
Merged
Vortex Layouts - Drivers #1914
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
robert3005
reviewed
Jan 13, 2025
vortex-layout/src/segments.rs
Outdated
| @@ -27,13 +27,19 @@ impl Deref for SegmentId { | |||
| pub trait AsyncSegmentReader: Send + Sync { | |||
| /// Attempt to get the data associated with a given segment ID. | |||
| /// | |||
| /// If the segment ID is not found, `None` is returned. | |||
| /// TODO(ngates): should this accept multiple segment IDs? | |||
There was a problem hiding this comment.
If the implementation is remote and invocation can have some non trivial constant overhead you want to take multiple but in most cases the answer to your question is "no"
robert3005
reviewed
Jan 13, 2025
Comment on lines
+287
to
+289
| // FIXME(ngates): how should we write into the segment cache? Feels like it should be | ||
| // non-blocking and on some other thread? | ||
| // segments.put(segment_id, buffer)?; |
There was a problem hiding this comment.
Why have this function if you don't put anything?
robert3005
reviewed
Jan 13, 2025
| } | ||
|
|
||
| /// TODO(ngates): outsource coalescing to a trait | ||
| fn coalesce(requests: Vec<FileSegmentRequest>) -> Vec<CoalescedSegmentRequest> { |
There was a problem hiding this comment.
why not take existing coalescing code? You can change the threshold upon which it choses to coalesce
robert3005
reviewed
Jan 13, 2025
| pub enum ExecutionMode { | ||
| /// Executes the tasks inline as part of polling the returned | ||
| /// [`vortex_array::stream::ArrayStream`]. In other words, uses the same runtime. | ||
| Inline, |
robert3005
reviewed
Jan 13, 2025
| } | ||
| .boxed() | ||
| } | ||
| Err(e) => futures::future::ready(Err(e)).boxed(), |
robert3005
reviewed
Jan 13, 2025
| let reader: Arc<dyn LayoutReader> = self | ||
| .layout | ||
| .reader(self.segments.reader(), self.ctx.clone())?; | ||
| pub fn scan(&self, scan: Arc<Scan>) -> VortexResult<impl ArrayStream + 'static + use<'_, I>> { |
There was a problem hiding this comment.
I'm confused by the lifetime signature here. How does use<...> change anything if the lifetime has to be 'static?
robert3005
reviewed
Jan 13, 2025
Comment on lines
+85
to
+86
| // Coalesce the segment requests to minimize the number of I/O operations. | ||
| .map(coalesce) |
There was a problem hiding this comment.
This just is a type wrapper?
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.
Introduces the idea of a
driveras a pluggable way of orchestrating I/O and CPU based work.An
ExecDrivertakes aStream<Future<T>>and returns aStream<T>. The abstraction means we can implement drivers that use the current runtime thread (Inline), spawn the work onto an explicit Runtime, or spawn the work onto a thread pool.The
IoDrivertakes a stream of segment requests and has to resolve them. For now, we have a FileIoDriver that assumes the bytes are laid out on disk (and therefore coalescing is desirable), so it pulls requests off the queue, coalesces them, then launches concurrent I/O requests to resolve them.The VortexFile remains generic over the I/O driver, meaning the send-ness of the resulting ArrayStream is inferred based on the send-ness of the I/O driver (in the case of FileDriver, that is the send-ness of the VortexReadAt).
Part of #1676