[
https://issues.apache.org/jira/browse/HADOOP-12620?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15083784#comment-15083784
]
Haohui Mai commented on HADOOP-12620:
-------------------------------------
I agree that the capabilities can be quite powerful. The real issue how it can
be done. There are some questions need to be answered:
(1) What is the semantic of update-in-place precisely when there are failures?
Is it atomic and transactional? What does the consistent model look like? What
are the semantics and durability guarantee look like? For example, what happens
if one of the DN in the pipeline is down? What will the reader see?
(2) Once you define the semantic, is the semantic / specification meaningful
and complete? Does it cover all the failure cases? How to evaluate and prove
there is no corner cases?
(3) How to implement the semantic in code? What is the approach you are taking?
Is it MVCC, distributed transaction or an ad-hoc solution tailored to HDFS?
So far we all agree that it is a useful capability. I don't think it require
more communications to establish it enables a number new use cases.
However, I don't see this is a complete solution without addressing Steve's
questions and all the questions above. It would be beneficial to have a design
doc and a working prototype to clarify the confusions.
> Advanced Hadoop Architecture (AHA) - Common
> -------------------------------------------
>
> Key: HADOOP-12620
> URL: https://issues.apache.org/jira/browse/HADOOP-12620
> Project: Hadoop Common
> Issue Type: New Feature
> Reporter: Dinesh S. Atreya
> Assignee: Dinesh S. Atreya
>
> Advance Hadoop Architecture (AHA) / Advance Hadoop Adaptabilities (AHA):
> See
> https://issues.apache.org/jira/browse/HADOOP-12620?focusedCommentId=15046300&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-15046300
> for more details.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
