[jira] [Closed] (JAMES-4141) Deduplication blob store: experiment Blake 3 instead of SHA
"Benoit Tellier (Jira)" <[email protected]>
| Newsgroups | gmane.comp.jakarta.james.devel |
|---|---|
| Message-ID | <[email protected]> |
[ https://issues.apache.org/jira/browse/JAMES-4141?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Benoit Tellier closed JAMES-4141.
---------------------------------
Resolution: Won't Fix
Experimentation is non-conclusive as the JVM fails to vectorize properly blake3.
It then executes worse the Sha256
> Deduplication blob store: experiment Blake 3 instead of SHA
> -----------------------------------------------------------
>
> Key: JAMES-4141
> URL: https://issues.apache.org/jira/browse/JAMES-4141
> Project: James Server
> Issue Type: Improvement
> Components: Blob
> Reporter: Tran Hong Quan
> Priority: Minor
> Time Spent: 1h 20m
> Remaining Estimate: 0h
>
> h2. Why?
> -> That is what Stallwart do.
> Blake 3 maintains security guaranties of SHA but can unlock an up to 10x speed improvement.
> There is a window of opportunity here: James spends ~5% of it's CPU hashing stuff for dedup so reducing its footprint is always a nice bonus. It can also allow reduce the impact of handling very big blobs.
> h2. What?
> Have a JVM property choosing if we are using SHA (conservative defaults) or want to enable Blake 3
> Implement Blake3 using [https://commons.apache.org/proper/commons-codec/apidocs/org/apache/commons/codec/digest/Blake3.html] (trivial)
--
This message was sent by Atlassian Jira
(v8.20.10#820010)