[SPARK-51711][ML][CONNECT] Memory based MLCache eviction policy #50530
+93
−17
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
|
Hi @WeichenXu123 , could you review this PR of changing entry count based MLCache eviction policy to memory based? |
zhengruifeng
reviewed
Apr 8, 2025
There was a problem hiding this comment.
let's make this jira a subtask of https://issues.apache.org/jira/browse/SPARK-51236
| buildConf("spark.connect.session.connectML.mlCache.maxSize") | ||
| .doc("Maximum size of the MLCache per session. The cache will evict the least recently" + | ||
| "used models if the size exceeds this limit. The size is in bytes.") | ||
| .version("4.0.0") |
There was a problem hiding this comment.
Suggested change
| .version("4.0.0") | |
| .version("4.1.0") |
I think this PR doesn't need to be included in 4.0.0
| buildConf("spark.connect.session.connectML.mlCache.timeout") | ||
| .doc("Timeout of models in MLCache. Models will be evicted from the cache if they are not " + | ||
| "used for this amount of time. The timeout is in minutes.") | ||
| .version("4.0.0") |
There was a problem hiding this comment.
Suggested change
| .version("4.0.0") | |
| .version("4.1.0") |
|
merged to master |
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Outdated suggestions cannot be applied.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.
What changes were proposed in this pull request?
Currently the ML Cache is limited by the number of cache entries (100 entries at this time), but it is not ideal because model size varies.
In this PR, we are updating the MLCache model eviction policy to be memory based, i.e. to evict old models if the total size of models is greater than a limit.
Besides, two new internal Spark confs are introduced:
spark.connect.session.connectML.mlCache.maxSize: Maximum size of the MLCache per session. The cache will evict the least recently used models if the size exceeds this limit.spark.connect.session.connectML.mlCache.timeout: Timeout of models in MLCache. Models will be evicted from the cache if they are not used for this amount of time.Why are the changes needed?
This improve the memory management of MLCache.
Does this PR introduce any user-facing change?
No.
How was this patch tested?
New test and existing tests.
Was this patch authored or co-authored using generative AI tooling?
No.