[
https://issues.apache.org/jira/browse/MNEMONIC-2?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15230543#comment-15230543
]
Wang, Gang commented on MNEMONIC-2:
-----------------------------------
Yes, this integration provides class MnemonicUnpooledDirectByteBuf and class
MnemonicUnpooledUnsafeDirectByteBuf, both of them implemented the interface
ByteBuff from Netty for Apache Arrow.
we embedded the Menmonic allocator into class PooledByteBufAllocatorL from
Apache Arrow instead of Netty allocator that basically use JVM internal to
allocate memory resource from main memory for DirectBuffer like SysMemAllocator
does from Mnemonic.
>From Mnemonic side, there is no new Allocator impls so far since the pmalloc
>and nvm serves general usage and there is no special request for allocator so
>far.
Yes I agree, we might introduce the Java collections of Apache Arrow to
Mnemonic if desired, Thanks.
> Apache Arrow Integration, Phase 1
> ---------------------------------
>
> Key: MNEMONIC-2
> URL: https://issues.apache.org/jira/browse/MNEMONIC-2
> Project: Mnemonic
> Issue Type: Improvement
> Reporter: Wang, Gang
> Assignee: Wang, Gang
> Labels: features
>
> The Apache Incubator Mnemonic provides a extendable mechanism for managing
> hybrid memory-like resources & their customized allocation services.
> Apache Arrow is powering columnar In-memory analytics. Theoretical, Apache
> Arrow is able to take advantages of Mnemonic and put its columnar dataset on
> various underlying memory-like devices e.g. SSD, NVMe, Off-heap.
> In this way, Apache Arrow could manage more columnar data for its
> applications and even customize its own allocation services according to
> their special data layouts for better performance as well as improve the
> generality and leverage Mnemonic coming new features if desired.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
