You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@hive.apache.org by "Sergey Shelukhin (JIRA)" <ji...@apache.org> on 2017/03/16 22:53:41 UTC
[jira] [Updated] (HIVE-16237) BuddyAllocator fragmentation -
long-term fix
[ https://issues.apache.org/jira/browse/HIVE-16237?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Sergey Shelukhin updated HIVE-16237:
------------------------------------
Description:
We were initially planning to write a different allocator with compaction, but now we are stuck with buddyallocator.
There are two approaches - compaction on top of buddy allocator (should be fairly simple actually, aside from indirection... luckily for the latter, we can take the buffer out in most places and we can reuse the existing locking-to-prevent-eviction mechanism to also prevent reallocation); or writing a better allocator. Probably the former is preferred, need to think about the latter.
was:The existing hacky code is not hacky enough.
> BuddyAllocator fragmentation - long-term fix
> --------------------------------------------
>
> Key: HIVE-16237
> URL: https://issues.apache.org/jira/browse/HIVE-16237
> Project: Hive
> Issue Type: Bug
> Reporter: Sergey Shelukhin
> Assignee: Sergey Shelukhin
>
> We were initially planning to write a different allocator with compaction, but now we are stuck with buddyallocator.
> There are two approaches - compaction on top of buddy allocator (should be fairly simple actually, aside from indirection... luckily for the latter, we can take the buffer out in most places and we can reuse the existing locking-to-prevent-eviction mechanism to also prevent reallocation); or writing a better allocator. Probably the former is preferred, need to think about the latter.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)