Accelerated DAG: Support dynamic resizing of shared memory channels #45323
+360
−94
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
jackhumphries
force-pushed
the
dynamic-size
branch
2 times, most recently
from
c4702ce
to
ced81ea
Compare
jackhumphries
force-pushed
the
dynamic-size
branch
from
e1b8fd0
to
2ff3953
Compare
jackhumphries
changed the title
|
This is ready for another pass. |
jackhumphries
force-pushed
the
dynamic-size
branch
from
a8ed927
to
2269eaf
Compare
stephanie-wang
added
the
@author-action-required
jackhumphries
force-pushed
the
dynamic-size
branch
from
8e2f5c4
to
017ea05
Compare
kevin85421
reviewed
May 29, 2024
jackhumphries
removed
the
@author-action-required
stephanie-wang
approved these changes
May 29, 2024
kevin85421
reviewed
May 30, 2024
|
|
||
| io_services_.push_back(std::make_unique<instrumented_io_context>()); | ||
| instrumented_io_context &io_service = *io_services_.back(); | ||
| io_works_.push_back(std::make_unique<boost::asio::io_service::work>(io_service)); |
There was a problem hiding this comment.
Use io_context instead? io_service seems to be deprecated boostorg/asio#110.
| std::vector<std::unique_ptr<boost::asio::io_service::work>> io_works_; | ||
| // Contexts in which the application looks for local changes to mutable objects and | ||
| // sends the changes to remote nodes via the network. | ||
| std::vector<std::unique_ptr<instrumented_io_context>> io_services_; |
| }); | ||
| } | ||
|
|
||
| void MutableObjectProvider::RunIOService() { | ||
| void MutableObjectProvider::RunIOService(instrumented_io_context &io_service) { |
There was a problem hiding this comment.
Suggested change
| void MutableObjectProvider::RunIOService(instrumented_io_context &io_service) { | |
| void MutableObjectProvider::RunIOContext(instrumented_io_context &io_context) { |
|
|
||
| io_services_.push_back(std::make_unique<instrumented_io_context>()); | ||
| instrumented_io_context &io_service = *io_services_.back(); | ||
| io_works_.push_back(std::make_unique<boost::asio::io_service::work>(io_service)); |
There was a problem hiding this comment.
Why do we need boost::asio::io_service::work here? Where do we call work.reset()?
| @@ -177,14 +183,19 @@ void MutableObjectProvider::PollWriterClosure( | |||
| object->GetData()->Size(), | |||
| object->GetMetadata()->Size(), | |||
There was a problem hiding this comment.
raylet_client_factory_: Creates a function for each object. This function waits for changes on the object and then sends those changes to a remote node via RPC.
The above comment is from experimental_mutable_object_provider.h. Can you explain how does reader->PushMutableObject "sends those changes to a remote node via RPC"? Thanks!
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.
Why are these changes needed?
In an accelerated DAG, all channels have their backing store specific upfront. However, this size may not be large enough for all objects that are communicated via channels, particularly since channels are reused across multiple DAG invocations.
This PR adds support for increasing the size of a shared memory channel backing store. Once the backing store size is increased, the size is maintained (i.e., it is not decreased) until it needs to be increased again.
Related issue number
Addresses one issue noted in #45597
Checks
git commit -s) in this PR.scripts/format.shto lint the changes in this PR.method in Tune, I've added it in
doc/source/tune/api/under thecorresponding
.rstfile.