Bind page-tail publication to proven recurrent boundaries - #35
Closed
FujitsuPolycom wants to merge 1 commit into
Closed
Bind page-tail publication to proven recurrent boundaries#35FujitsuPolycom wants to merge 1 commit into
FujitsuPolycom wants to merge 1 commit into
Conversation
SparkCache opts into vLLM recurrent-boundary hand-offs and accepts them only when request, group, token boundary, recurrent topology, and non-null physical block agree. Every aligned recurrent group must have exactly one proven block before a store plan is created. Absent, incomplete, contradictory, or preemption-stale metadata cancels publication. SparkCache never scans arithmetic, running, or speculative slots for a substitute. GPU-free coverage exercises the 6,912-token boundary with seven DFlash verification slots through SparkContextCacheConnector and ManifestStore, including malformed coverage and request-lifetime cleanup. The exact GLM vLLM lease contract accepts and requires the recurrent-boundary runtime postimages and SchedulerOutput interface. Cache namespace impact: none. CacheIdentity values, digest salts, chunk geometry, manifest schemas, and page-delta wire bytes are unchanged. Deployable SparkCache source SHA-256: 01cc59bf2c45af60f02813b771a484e65829e4f464eb27aa1756ca7067c82c9f. Validation: python -m pytest sparkcache -q (752 passed, 7 skipped); python -m pytest deploy -q (108 passed, 1 skipped); python -m ruff check .; source-attestation tests (2 passed). One concurrent validation run exceeded an existing sub-millisecond timing assertion; the isolated test and the following full SparkCache run passed.
This was referenced Aug 30, 2026
Owner
Author
This file contains hidden or 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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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.
Resulting behavior
Status: implemented and GPU-free tested; live model-serving qualification is pending.
For an aligned recurrent KV group, vLLM may retain the replay-boundary page outside the advancing request block table. The scheduler now supplies SparkCache with a connector-specific
recurrent_boundary_blockshandoff. Each entry identifies one request, KV group, physical block, and exact token boundary.SparkCache accepts the retained block only when request identity, group identity, boundary identity, recurrent topology, and worker ownership all agree. Missing, partial, duplicate, null, stale, wrong-group, or wrong-boundary metadata cancels publication for that request. SparkCache never searches later running state or DFlash verification slots for a plausible block.
Overlapping producer steps defer block recycling until worker snapshot completion. Request and preemption cleanup release the retained blocks. Connectors that do not advertise the capability retain existing behavior.
This branch composes the recurrent-boundary contract with the 64 MiB page-delta macro objects in the parent branch. Version 1 page-delta manifests remain readable; version 2 reduces the 1,575,821,491-byte test scale from 1,024 physical objects to at most 24.
Compatibility
Cache identity values, digest salts, 256-token logical geometry, page-tail namespace, and the vLLM connector interface for producers without this capability are unchanged. Cache namespace impact: none.
The combined deployable SparkCache source SHA-256 is
88633ef676b4dfe258a6fa9b788ddeb22cad68349d0cae0c503ee404d1724f7b. The exact vLLM recurrent-boundary contract SHA-256 remains45d7a92b38b836a4f829f02df85e339cfeea860e1080e4663a8340af6c125125.Validation
This draft is stacked on page-delta macro-object PR #33. It provides the combined source intended for an image build and live qualification; it does not merge, publish an image, or change the active service. It supersedes the SparkCache side of PR #31 only after review.