You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@hbase.apache.org by "Vladimir Rodionov (Jira)" <ji...@apache.org> on 2026/04/23 00:04:00 UTC

[jira] [Updated] (HBASE-30109) Add CarrotCache-based CacheEngine implementation for pluggable block cache framework

     [ https://issues.apache.org/jira/browse/HBASE-30109?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Vladimir Rodionov updated HBASE-30109:
--------------------------------------
    Description: 
h2. Description

Add a *CarrotCache*-based *CacheEngine* implementation for the pluggable block cache framework.

This change introduces CarrotCache as an alternative storage backend for HBase block caching. The initial goal is to integrate it into the new cache architecture without changing HBase block format semantics or read/write path behavior.

CarrotCache is a good candidate for this role because it provides:

* off-heap storage
* low metadata overhead
* optional compression-aware storage
* high cache density for large-capacity deployments

h3. Scope

* Implement a new CacheEngine backed by CarrotCache
* Integrate the new engine with:
** CacheTopology
** CachePlacementPolicy
** CacheAccessService
* Support use of CarrotCache as an L2 cache engine in the initial version
* Store HBase-compatible cached block payloads without introducing new block format semantics
* Expose capacity, usage, invalidation, and statistics through the CacheEngine contract

h3. Notes

* This ticket depends on prior introduction of:
** CacheEngine
** CacheTopology
** CachePlacementPolicy
** CacheAccessService
* No changes to HBase read/write semantics are intended in the initial integration
* The first implementation should prioritize:
** correctness
** compatibility
** low-risk integration
over more aggressive optimizations

* This ticket does *not* include:
** changes to HFile block format
** policy changes specific to CarrotCache
** advanced representation optimizations beyond what is required for initial integration

* Expected outcome:
** an alternative cache engine with lower metadata overhead
** improved flexibility for large-capacity cache deployments
** a concrete validation of the pluggable cache architecture

> Add CarrotCache-based CacheEngine implementation for pluggable block cache framework
> ------------------------------------------------------------------------------------
>
>                 Key: HBASE-30109
>                 URL: https://issues.apache.org/jira/browse/HBASE-30109
>             Project: HBase
>          Issue Type: New Feature
>          Components: BlockCache, Performance
>            Reporter: Vladimir Rodionov
>            Assignee: Vladimir Rodionov
>            Priority: Major
>
> h2. Description
> Add a *CarrotCache*-based *CacheEngine* implementation for the pluggable block cache framework.
> This change introduces CarrotCache as an alternative storage backend for HBase block caching. The initial goal is to integrate it into the new cache architecture without changing HBase block format semantics or read/write path behavior.
> CarrotCache is a good candidate for this role because it provides:
> * off-heap storage
> * low metadata overhead
> * optional compression-aware storage
> * high cache density for large-capacity deployments
> h3. Scope
> * Implement a new CacheEngine backed by CarrotCache
> * Integrate the new engine with:
> ** CacheTopology
> ** CachePlacementPolicy
> ** CacheAccessService
> * Support use of CarrotCache as an L2 cache engine in the initial version
> * Store HBase-compatible cached block payloads without introducing new block format semantics
> * Expose capacity, usage, invalidation, and statistics through the CacheEngine contract
> h3. Notes
> * This ticket depends on prior introduction of:
> ** CacheEngine
> ** CacheTopology
> ** CachePlacementPolicy
> ** CacheAccessService
> * No changes to HBase read/write semantics are intended in the initial integration
> * The first implementation should prioritize:
> ** correctness
> ** compatibility
> ** low-risk integration
> over more aggressive optimizations
> * This ticket does *not* include:
> ** changes to HFile block format
> ** policy changes specific to CarrotCache
> ** advanced representation optimizations beyond what is required for initial integration
> * Expected outcome:
> ** an alternative cache engine with lower metadata overhead
> ** improved flexibility for large-capacity cache deployments
> ** a concrete validation of the pluggable cache architecture



--
This message was sent by Atlassian Jira
(v8.20.10#820010)