[jira] [Commented] (HDFS-8833) Erasure coding: store EC schema and cell size in INodeFile and eliminate notion of EC zones

["Zhe Zhang (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/HDFS-8833?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14653161#comment-14653161 ] 

Zhe Zhang commented on HDFS-8833:
---------------------------------

Thanks Andrew for the suggestion. I think {{*-on-create}} flags is a good solution to the future compatibility concern.

[~walter.k.su] also mentioned some ideas in an offline discussion.

To summarize, below is the proposed design for this change:

# ErasureCodingPolicy table
#* Create {{ErasureCodingPolicy}} class, with {{ECSchema}} and {{cellSize}}.
#* Create {{ErasureCodingPolicySuite}} class to manage a table of supported policies (or extend {{ErasureCodingSchemaManager}}). Something like:
{code}
0: RS-6-3 (schema), 64KB (cellSize)
1: RS-6-3 (schema), 128KB (cellSize)
2: RS-10-4 (schema), 1MB (cellSize)
{code}
#* [follow-on] Allow customize policies stored in XAttr
# File header change
#* Remove {{isStriped}} from {{INodeFile}} header and reduce replication factor to 6 bits.
{code}
  /** 
   * Bit format:
   * [4-bit storagePolicyID][6-bit erasureCodingPolicy]
   * [6-bit replication][48-bit preferredBlockSize]
   */
{code}
#* Store ID of ECPolicy with 6 bits in header -- 64 policies allowed
#* The ECPolicy is *always set* when creating a file, taking value from its ancestors; {{0}} can be used to represent contiguous layout.
#* [follow-on] Add {{inherit-on-create}} flag as Andrew suggested above
# Directory XAttr change
#* Directory's ECPolicy XAttr can be empty, indicating the ECPolicy is the same as ancestor. Otherwise its own XAttr determines the policy for newly created files under the directory.
# Renaming
#* A renamed file keeps the ECPolicy in its header.
#* Therefore, a directory can have files with different ECPolicies.
#* Conversion not explicitly support. If needed a file can be converted by cp+rm.
#* When renamed, a directory carries over its ECPolicy if it's set (XAttr non-empty). Otherwise its XAttr remains empty (and newly created files under the moved directory will use policy from the new ancestors). 

Questions / comments are very welcome.

> Erasure coding: store EC schema and cell size in INodeFile and eliminate notion of EC zones
> -------------------------------------------------------------------------------------------
>
>                 Key: HDFS-8833
>                 URL: https://issues.apache.org/jira/browse/HDFS-8833
>             Project: Hadoop HDFS
>          Issue Type: Sub-task
>          Components: namenode
>    Affects Versions: HDFS-7285
>            Reporter: Zhe Zhang
>            Assignee: Zhe Zhang
>
> We have [discussed | https://issues.apache.org/jira/browse/HDFS-7285?focusedCommentId=14357754&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-14357754] storing EC schema with files instead of EC zones and recently revisited the discussion under HDFS-8059.
> As a recap, the _zone_ concept has severe limitations including renaming and nested configuration. Those limitations are valid in encryption for security reasons and it doesn't make sense to carry them over in EC.
> This JIRA aims to store EC schema and cell size on {{INodeFile}} level. For simplicity, we should first implement it as an xattr and consider memory optimizations (such as moving it to file header) as a follow-on. We should also disable changing EC policy on a non-empty file / dir in the first phase.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)