refactor(inkless:consume): replace ByteRange.coalesce with BoundingRangeAlignment strategy by jeqo · Pull Request #532 · aiven/inkless

jeqo · 2026-03-10T13:29:24Z

to reduce read amplification

FetchPlanner.createFetchRequests() unconditionally aligned all byte ranges to fixed blocks (e.g. 16 MiB) before the hot/cold path decision. For the cold path, the cache is bypassed entirely, so alignment padding is pure read amplification with zero benefit — a 10 KB batch could trigger a 16 MiB fetch.

Move the hot/cold determination into the planning phase so that:

Hot path: applies block alignment as before (cache benefit)
Cold path: uses coalesced exact batch ranges (minimal bytes fetched)

Add ByteRange.coalesce() to merge overlapping/adjacent ranges, and extract isLagging() helper to deduplicate the threshold check between createFetchRequests() and submitSingleRequest().

Copilot

Pull request overview

This PR reduces read amplification for lagging consumers by choosing the byte-range strategy during planning: hot-path requests remain block-aligned for cache efficiency, while cold-path requests bypass cache and fetch coalesced exact ranges.

Changes:

Update FetchPlanner to select aligned vs coalesced ranges based on lagging/hot determination during planning.
Add ByteRange.coalesce() to merge overlapping/adjacent ranges into a minimal set.
Extend unit tests to validate cold-path planning behavior and ByteRange.coalesce() semantics.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 4 comments.

File	Description
`storage/inkless/src/main/java/io/aiven/inkless/consume/FetchPlanner.java`	Chooses coalescing vs alignment during planning and extracts `isLagging()` helper used by both planning and submission.
`storage/inkless/src/main/java/io/aiven/inkless/common/ByteRange.java`	Adds `coalesce(List<ByteRange>)` utility to merge overlapping/adjacent ranges.
`storage/inkless/src/test/java/io/aiven/inkless/consume/FetchPlannerTest.java`	Adds planning-focused tests to ensure cold path uses exact/coalesced ranges and hot path stays aligned.
`storage/inkless/src/test/java/io/aiven/inkless/common/ByteRangeTest.java`	Adds test coverage for `ByteRange.coalesce()` behavior across overlap/adjacency/ordering cases.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

storage/inkless/src/main/java/io/aiven/inkless/consume/FetchPlanner.java

storage/inkless/src/test/java/io/aiven/inkless/consume/FetchPlannerTest.java

storage/inkless/src/test/java/io/aiven/inkless/common/ByteRangeTest.java

Copilot

Pull request overview

Copilot reviewed 5 out of 5 changed files in this pull request and generated 1 comment.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

storage/inkless/src/main/java/io/aiven/inkless/common/ByteRange.java

Copilot

Pull request overview

Copilot reviewed 5 out of 5 changed files in this pull request and generated no new comments.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

gharris1727 · 2026-03-11T15:12:43Z

storage/inkless/src/main/java/io/aiven/inkless/common/ByteRange.java

+     * @param ranges the byte ranges to coalesce (may be null or empty)
+     * @return an unmodifiable set of merged, non-overlapping byte ranges in offset order
+     */
+    public static Set<ByteRange> coalesce(List<ByteRange> ranges) {


This is the signature of KeyAlignment, which makes me think this is inappropriate complexity to add as a static method.

gharris1727 · 2026-03-11T15:15:18Z

storage/inkless/src/main/java/io/aiven/inkless/common/ByteRange.java

+                // Overlapping or adjacent — extend
+                currentEnd = Math.max(currentEnd, nextEnd);
+            } else {
+                // Gap — emit current and start new


If there are gaps, they are expected to be small, and making multiple requests to skip around that gap would be worse than just throwing around the data.
This routine shouldn't do this sort/match end-to-end algorithm, it should compute the minimum start and maximum end offset.

Good call! I have updated the implementation to go along this approach. PTAL

storage/inkless/src/main/java/io/aiven/inkless/cache/BoundingRangeAlignment.java

storage/inkless/src/test/java/io/aiven/inkless/cache/BoundingRangeAlignmentTest.java

…ngeAlignment strategy Reuse KeyAlignmentStrategy interface for cold-path range computation instead of a static method on ByteRange. One bounding range per object minimizes HTTP requests at the cost of reading small interleaved gaps.

…ngeAlignment strategy (#532) Reuse KeyAlignmentStrategy interface for cold-path range computation instead of a static method on ByteRange. One bounding range per object minimizes HTTP requests at the cost of reading small interleaved gaps. (cherry picked from commit 069cce8)

…ngeAlignment strategy (#532) Reuse KeyAlignmentStrategy interface for cold-path range computation instead of a static method on ByteRange. One bounding range per object minimizes HTTP requests at the cost of reading small interleaved gaps.

…ngeAlignment strategy (#532) Reuse KeyAlignmentStrategy interface for cold-path range computation instead of a static method on ByteRange. One bounding range per object minimizes HTTP requests at the cost of reading small interleaved gaps. (cherry picked from commit 069cce8)

jeqo marked this pull request as ready for review March 10, 2026 14:43

jeqo requested a review from Copilot March 10, 2026 14:43

Copilot started reviewing on behalf of jeqo March 10, 2026 14:44 View session

Copilot AI reviewed Mar 10, 2026

View reviewed changes

jeqo force-pushed the jeqo/reduce-fetch-read-amplification branch from 28f05c8 to a5ecdf7 Compare March 10, 2026 16:41

jeqo requested a review from Copilot March 10, 2026 16:42

Copilot started reviewing on behalf of jeqo March 10, 2026 16:42 View session

Copilot AI reviewed Mar 10, 2026

View reviewed changes

storage/inkless/src/main/java/io/aiven/inkless/common/ByteRange.java Outdated Show resolved Hide resolved

jeqo force-pushed the jeqo/reduce-fetch-read-amplification branch from a5ecdf7 to 583b55a Compare March 10, 2026 16:58

jeqo requested a review from Copilot March 10, 2026 16:58

Copilot started reviewing on behalf of jeqo March 10, 2026 16:59 View session

Copilot AI reviewed Mar 10, 2026

View reviewed changes

jeqo force-pushed the jeqo/reduce-fetch-read-amplification branch from 583b55a to 2019a0b Compare March 10, 2026 17:26

gharris1727 reviewed Mar 11, 2026

View reviewed changes

jeqo force-pushed the jeqo/reduce-fetch-read-amplification branch from 2019a0b to f716ab6 Compare March 11, 2026 17:20

jeqo changed the title ~~feat(inkless:consume): skip block alignment for lagging consumers~~ refactor(inkless:consume): replace ByteRange.coalesce with BoundingRangeAlignment strategy Mar 11, 2026

jeqo force-pushed the jeqo/reduce-fetch-read-amplification branch from f716ab6 to d709c1b Compare March 11, 2026 18:41

jeqo requested review from gharris1727 and jjaakola-aiven March 11, 2026 19:17

viktorsomogyi reviewed Mar 20, 2026

View reviewed changes

storage/inkless/src/main/java/io/aiven/inkless/cache/BoundingRangeAlignment.java Outdated Show resolved Hide resolved

storage/inkless/src/test/java/io/aiven/inkless/cache/BoundingRangeAlignmentTest.java Outdated Show resolved Hide resolved

jeqo force-pushed the jeqo/reduce-fetch-read-amplification branch from d709c1b to 83f2f9f Compare March 20, 2026 13:35

viktorsomogyi approved these changes Mar 20, 2026

View reviewed changes

jeqo merged commit 069cce8 into main Mar 20, 2026
4 checks passed

jeqo deleted the jeqo/reduce-fetch-read-amplification branch March 20, 2026 14:22

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

refactor(inkless:consume): replace ByteRange.coalesce with BoundingRangeAlignment strategy#532

refactor(inkless:consume): replace ByteRange.coalesce with BoundingRangeAlignment strategy#532
jeqo merged 1 commit intomainfrom
jeqo/reduce-fetch-read-amplification

jeqo commented Mar 10, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Copilot AI left a comment

Uh oh!

Uh oh!

Copilot AI left a comment

Uh oh!

gharris1727 Mar 11, 2026

Uh oh!

gharris1727 Mar 11, 2026

Uh oh!

jeqo Mar 11, 2026

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

Conversation

jeqo commented Mar 10, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

gharris1727 Mar 11, 2026

Choose a reason for hiding this comment

Uh oh!

gharris1727 Mar 11, 2026

Choose a reason for hiding this comment

Uh oh!

jeqo Mar 11, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants