[ 
https://issues.apache.org/jira/browse/HDFS-11138?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15675319#comment-15675319
 ] 

Chen Liang commented on HDFS-11138:
-----------------------------------

Thanks [~anu] for the comments! Again here are some thoughts I had (mainly 
related to 3 & 4 & 6). And will address the rest in the next path.

I know that keeping a map AND a list in StorageManager is not quite clean. The 
idea was that I wanted to allow two ways of accessing containers:
1. access them from the first to the last, based on their order index - which 
requires a list
2. access a specific a container with a given containerID - which requires a map
This was my very first design when writing the class. 

So far, we access all containers at once on mount time by returning all the 
containers along with their order, in which case a list is sufficient. I wanted 
to leave the possibility for accessing a container by key on server side, but I 
can't really think of a use case for the time being... Do you think this 
access-by-key is not worth keeping? If so, I'll just go remove it.

> Block Storage : add block storage server
> ----------------------------------------
>
>                 Key: HDFS-11138
>                 URL: https://issues.apache.org/jira/browse/HDFS-11138
>             Project: Hadoop HDFS
>          Issue Type: Sub-task
>          Components: hdfs
>            Reporter: Chen Liang
>            Assignee: Chen Liang
>         Attachments: HDFS-11138-HDFS-7240.001.patch, 
> HDFS-11138-HDFS-7240.002.patch, HDFS-11138-HDFS-7240.003.patch, 
> HDFS-11138-HDFS-7240.004.patch
>
>
> This JIRA adds the skeleton for server side code, which is one of the core 
> components in block storage. For performance concerns, the server does not 
> handle any actual read/write operation but serving primarily as a meta data 
> server. It provides four APIs:
> # create volume : which will call into underlying container layer to allocate 
> containers
> # delete volume : delete a specific volume (as well as its containers)
> # info volume : return information of a specific volume
> # list volume : list all volumes
> Note that this is still subject to potentially major changes. Features such 
> as persistence are missing.



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

---------------------------------------------------------------------
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org

Reply via email to